Twenty-one things this archive got wrong, and what fixed them

What Changed

An archive that never contradicts itself is not being careful; it is not checking. This is the list of claims made here and then withdrawn, kept in the open because the reasoning that failed is as useful as the reasoning that held.

Every page here carries a confidence label, and a label is only honest if it can be lowered. Most of what follows is this archive becoming less certain, not more: a house number demoted from an identifier to a kin-group, a resolved name reopened, a culprit un-named. One was caught by the reader — the search had been folding surnames and not given names, and was finding 42 of 88 Ivans. One was caught by sweeping the built site rather than the source, which is a lesson about where to look.

13 September 2026

1. Šojat is a family. I called it a by-name.

This archive said	Recorded in the register as the by-name the tree attaches to one Blažević household — read as a nadimak, like Papa on the Antići.
The record says	The Šojati are a Krivi Put family of fifteen people with their own dates and their own place — Marko Šojat, 19 April 1801 to 4 February 1871, five dated children, and a valley called Sojat Dol. They married the Blaževići and Prpići nine times.
What it cost	The mechanism was right and the direction was backwards. Blažević (Sojat) is almost certainly a household known by the family it married — which is how a nadimak is born, not what one is. Who they married →

This archive said	That the tree held 232 Žubrinići in 94 pieces, and that joining them was the largest single piece of work outstanding in this archive.
The record says	The surname fold split on ( but not on /. So Žubrinić / Xubrinich and Prpic Perpic — one surname written several ways in a single field — never matched the surname list. The real figure is 422, and fixing it brought 214 people into the archive, including Elias Ilija Žubrinić and Casparus Žubrinić, who are the subjects of two of this archive's own open questions.
What it cost	Two published numbers, and the conclusion drawn from them. Counted properly there are 90 loose ends, not 94 — and 79 of them are bare names with no date and no place. So the work is not large; it is impossible from here, and the register is the only way through. The fragments, counted →

This archive said	Nothing — this had never been noticed. The tree looked more joined than it is.
The record says	Eighteen nodes whose entire given name is a single letter — M, J, T — each with a parent called “Žubrinić Zubrinix Xubrinic Zub” and each holding as children everyone whose name starts with that letter. One has 26 children.
What it cost	92 people are attached to the letter their name begins with and to nothing else. The sorting-bucket filter missed them because a bucket is normally detected by words like to be sorted, and these are named A through W. Read as parentage — which is what a naive count does — it makes the Žubrinići look like a family.

This archive said	That Smokvica Krmpotska kbr. 114 identified one family, and so placed Juraj Blažević in Toma's household — nearly naming his father.
The record says	A control document, the 1890 baptism of Frane Blažević, puts Pavao Blažević, son of Luka, at 114 as well. Luka and Toma's father Anton are both sons of Ilija, so Pavao and Toma are first cousins and the number covers an extended-family compound — a zadruga.
What it cost	Question 1b was downgraded from critical to substantive. The house places Juraj in Ilija's kin-group; it cannot name his father.

This archive said	That he should be looked for in the Austro-Hungarian Verlustlisten, reasoning from his age — thirty-two in the second year of the war.
The record says	The tree places his death at Senj and gives his occupation as gostioničar, bankar. A man who dies at home in a settled trade is not a casualty-list entry.
What it cost	The right book is the Senj death register for 1915. The reasoning was plausible and it was still wrong: an age is not a cause.

This archive said	A block on Karlobag's merchant milieu placed on the Žubrinić page.
The record says	It arrives through Matija Pilipić, ahnentafel 13 — the Papić line. Moved.
What it cost	Nothing to the evidence; everything to whether a reader can follow the line it belongs to.

This archive said	Both drafted to Državni arhiv u Gospiću, on the assumption that Lika-Senj county means Gospić.
The record says	Gospić's remit is Lika-Senj “osim gradova Senja i Novalje” — except the towns of Senj and Novalja. And Smokvica is in Primorje-Gorski Kotar, not Lika-Senj at all. Both belong to Rijeka, Senj branch.
What it cost	Had they been sent, two archives would have spent time forwarding them. Jurisdiction is now written down before anything is posted.

This archive said	That folding Žubrinić / Xubrinich / Zubrinich together was enough to make a person findable.
The record says	It was not. Ivan found 42 of 88 Ivans, because the rest are entered Joannes, Joannis, Juan or Ive. 268 canonical given names, 34 of them under more than one spelling.
What it cost	Caught by the reader, not by this archive. Both surnames and given names now fold — and the split by sex was a second correction on top, after Franjo and Franciska were wrongly merged.

This archive said	That Josephum, Joannis and Mariam were one person working through Selce and Vinodol entries and lifting the word after baptizavi.
The record says	Three scans, two parishes, three priests, three decades apart — the priests wrote the oblique forms themselves. Michaël Kombol writes nomine + accusative for every child on the 1806 page.
What it cost	The artefact is real; the culprit was invented. It also turned the ending into a useful signal — accusative means the entry is the person's own, genitive means they were only named as somebody's parent.

This archive said	That Luca “Papić or Perpic” was settled, because her siblings are recorded Papić.
The record says	Her father is a record called “Senj — Unknown To be Sorted Papić”. Those siblings were grouped by whoever compiled the tree, not by a register.
What it cost	The tooling now reuses the same sorting-bucket filter the rest of the build uses, so a bucket can never resolve a name. Marked open.

This archive said	Mariam Gerkacs (Gergacs), ahnentafel 55, following the tree's ordering.
The record says	The Karlobag register of August 1789 spells it Gergacs. The tree leads with the secondary form.
What it cost	Small, and worth doing: the register's spelling is the one an archivist will search on.

This archive said	Question 5c, published this morning, said four key scans had been downloaded from MyHeritage at the wrong size and should simply be re-fetched — no archive, no letter, no travel.
The record says	Testing it took two minutes and disproved it. The MyHeritage CDN returns exactly 960 pixels for the 1885 marriage, so the file held in the tree is itself 960 pixels. What separates the good scans is where they came from: every 4,500-pixel document carries a FamilySearch image identifier in its filename, and no 960-pixel one does.
What it cost	The wrong fix. Re-fetching would have returned the same four thumbnails and the question would have looked answered. The real action is to re-source three entries from FamilySearch — and the registers and pages for all three are already known.

This archive said	Five items in the media library shown as photographs of Michael Žubrinić (b. 1811), Maria Catharina Orešković (1835), Josephum Antić (1806), Maria Jelličić (1809) and Georgius Antić Papić (1838).
The record says	They are stock question-mark placeholder graphics, uploaded to the tree for people who have no picture. One still carries a visible Adobe Stock watermark. The dates alone should have caught it: a photograph of a man born in 1806 would be remarkable.
What it cost	They were on the pictures page and on two person pages, captioned as those people. The media build now honours an excluded classification and drops them. A sixth was reclassified rather than dropped — a picture captioned “Ljubomir Blazevic” is the family gravestone, not a portrait.

This archive said	Otočac shown with 60 Žubrinić records; four Mrzli Dol records filed under a place called “#6 Mrzli Dol”.
The record says	Otocac Croatia is Otočac — 68 records. The house-number pattern missed `#` and `kbr.`, hiding houses 9 and 30 entirely.
What it cost	Every count on the map page was low. Fixed when the tool was widened from one surname to four.

What	The tree	The record
Corrected Baptism date	19 October 1886	17 October 1886	→
Noted Birth date	1 March 1892 — correct	The gravestone at Senj says 1891	→
Disputed Attachment to this family	Wife of Michael Žubrinić; mother of Nicolaus (b. 24 January 1834)	Born 8 September 1835 — twenty months after her supposed son	→
upgraded Parentage of his son Ljubomir	Asserted, with no document — this archive called it Disputed	Documented: Senj marriages, 8 December 1920, page 366 entry 41	→
Confirmed Birth date	14 March 1806	14 March 1806 — confirmed	→
Noted The name Josephum	Josephum	Josephus — and the priest wrote the accusative himself	→
Corrected Surname	Gerkacs (Gergacs)	Gergacs	→
Noted Birth date	11 August 1789	August 1789, born the day before baptism — consistent	→
Noted Baptism entry	Born 22 April 1777, Karlobag	April 1777, Karlobag, born the day before — consistent	→
Corrected She is not this child's mother	Mother of Nicolaus Žubrinić, b. 24 January 1834	Born 8 September 1835 — nineteen months after him	→
Confirmed Baptism	24 January 1834, Otočac	24 January 1834, Otočac — confirmed	→
Noted Two wives, or two men	Married to Maria Catharina Orešković	1834: Maria nata Oreskovich · 1845: Maria nata Marković	→
Corrected Birth date	17 May 1884	12 May 1884	→
Confirmed Birth date, and her parents' house	25 October 1884	25 October 1884 — confirmed	→
Disputed His parents	Michael “Mate” Žubrinić and Maria Catharina Orešković — the same couple as Nicolaus, b. 1834	Mate Žubrinić and Marija r. Orešković, both dead by 1910 — but a different couple	→
Confirmed Birth date	19 October 1885	19 October 1885 — confirmed	→
Confirmed Birth date	23 October 1888	23 October 1888 — confirmed	→
Confirmed Death, cause and burial	26 July 1907, Senj — no cause, no burial detail	26 July 1907, Senj — tuberculosis, aged 69, buried sv. Vid on 28 July	→
Corrected Death date	Before 1850	After 1884, and before 26 July 1907	→
Confirmed Baptism, and her father's occupation	Born 1838, Karlobag	Baptised 18 September 1838, Karlobag — confirmed	→
Noted Burial	Not recorded	Senj — the stone reads BLAŽEVIĆ TEREZA ✻1867 †1931	→
Disputed This baptism is not his	Born 12 January 1779 — the tree's own note adds “(Maybe)”	A real Karlobag baptism of that date, but of a different man	→
Disputed Attached on a baptism that is not his son's	Father of Caetano Bacchi	Named only in the 1779 entry, which cannot be ahn 214's	→
Confirmed Birth	March 1781, Karlobag	Born 11 March 1781 — confirmed	→
Confirmed Baptism	11 March 1768, Karlobag	Baptised 11 March 1768, born 10 March — confirmed	→
Confirmed Birth and baptism	Born 12 May 1859, baptised 15 May 1859, Krivi Put 136	Confirmed exactly — and the household is entered Confiniarii	→
Confirmed Marriage and trade	Married Antonija Prpić; trhonoša, bremenar, gostioničar	Married 3 May 1885 at Senj, aged 27, entered bremenar	→
Noted Still no surname, after two documents	Lucia “Luca”, no surname	Two registers name her and neither gives one	→
Disputed A child born after his death	Died 20 March 1954; Vladimira Boras born 1960	Not reconciled	→
Disputed A child born 74 years after his death	Died 1888; Georgius Prpić born 1962	Not reconciled	→
Disputed A child born after his death	Died 1926; Blazenka Boras born 1930	Not reconciled	→
Disputed A child born after his death	Died 1845; Magdalena Žubrinić born 1847	Not reconciled	→
Disputed Died before she was born	Born 11 February 1847, died 4 August 1846	Not reconciled	→
Disputed A mother at ten	Born 21 June 1812; Lucas Ugrin Adamich born 1822	Not reconciled	→
Noted A father in the same year as his son	Born before 1769; Stephanus Žubrinić born 1769	An artefact of the hedge, not a contradiction	→
Corrected Birthplace — the county is wrong	Smokvica, Krmpote, Općina Smokvica, Dubrovačko-neretvanska županija	Smokvica Krmpotska, Općina Novi Vinodolski, Primorje-Gorski Kotar	→
Corrected Birthplace — the county is wrong	Senj, Split-Dalmatia, Croatia	Senj, Lika-Senj County, Croatia	→
Confirmed Death, cause and age	Born 1902, died 1903	Died 3 June 1903 aged 1 year 8 months — scarlatina with diphtheria	→
Corrected Her father	Mother Katarina “Kate” Matijavic; no father recorded	Franjo Papić, nadničar	→
Noted Where her death is registered	Died 7 September 1971	Confirmed — and the civil entry is cited in her baptism margin	→

This archive said	That the 1918 marriage entry for Ana Papić gave her mother as Antonija r. Papić — a clerk confusing the two family names.
The record says	Enlarged, it plainly reads Prpić. There was no clerk's error.
What it cost	Nothing, because it was caught before it was published as a finding. Recorded because a correction that never reached the page is still a correction.

This archive said	Open questions rewrote `href='/TheBlazevic/…'` in the question body but not in the “what would settle it” block.
The record says	Both fields go through one helper. Two of the five had been dead since the page was written.
What it cost	A reader clicking the most actionable part of the page got a 404. Found by sweeping the built site rather than the source.

This archive said	That the wrong-attachment problem was contained: a real Maria Catharina Orešković of Dubrava had been married to the wrong man, and the rest of the line stood.
The record says	Reading the 1845 Otočac baptism of Josephus gives the mother as Maria Žubrinić nata Marković, where the 1834 baptism of Nicolaus gives Maria nata Orešković. Either the first wife died and he married a second Maria, or there are two men of this name — there are 68 Žubrinić records at Otočac.
What it cost	The archive had already established that Maria Catharina was the wrong woman — that reversal is set out on the Frontier page and is not new here. What is new is that the husband is now uncertain too, which is a larger problem than the one it replaces.

This archive said	Caetano Bacchi, ahnentafel 214, born 12 January 1779 — carried from the tree, whose own note reads “(Maybe)”.
The record says	The 1779 Karlobag entry is real, but ahn 214's daughter died in 1795 and her daughter was born 11 March 1781. Two years cannot hold two generations. The right man belongs around 1730–1745.
What it cost	Two more ancestors go with it: Angelo Bacchi and Catharina are in this pedigree only because that entry names them. Marked disputed, not deleted — the wrong join is the only pointer to where the right man should be sought.

This archive said	The 1777 Uroda godfather given as Pro-Colonellus Joannes Ballinich* de Zirfeld*.
The record says	The same surname recurs in a February 1779 entry where the words Joës Bapta sit on the same line — giving a clean two-bowled capital B to compare against. The surname's initial is a single looped ascender with a leg, and plainly not that letter. It is Hallinich de Zirfeld.
What it cost	Small in itself, and the method is the point: a contested letter was settled by finding the candidate letter elsewhere in the same hand, rather than by staring harder at the word.

This archive said	1800-1830 Birth — Selce Papić (Papa), two more like it, and 1700 Birth — Selce Papić, all with person pages and children attached.
The record says	They are research scaffolding, not people. The build has always excluded such records — the code comment says publishing them “would be a lie” — but the filter did not recognise a name opening with a year, nor Brothers of Casparus (It Seems).
What it cost	Four people who never existed had pages, and were parents in other people's family lists. The person count fell by four. Found not by reading but by the new self-consistency check: a bucket dated 1800-1830 cannot father someone born in 1824.

This archive said	Tereza Pećanić, ahnentafel 9, carried through from the tree as died before 1850 — published without anyone checking it against the rest of her own record.
The record says	The same record marries her to Toma on 21 November 1858 and gives her children in 1860, 1880, 1882 and 1884. Toma's death entry of 26 July 1907 calls him udovac iza pok. Tereze — widower of the late Tereza. So she was alive in 1884 and dead by July 1907.
What it cost	Nothing downstream — no date was computed from it. But it sat on a direct ancestor's page for the life of this archive, and it is the kind of error that needs no register to catch: her own children postdate her death by up to thirty-four years. The lesson is that internal consistency was never checked, only external agreement.

What Changed

1. Šojat is a family. I called it a by-name.

2. A bracket and a slash cost this archive 214 people.

3. Ninety-two Žubrinići were held together by an alphabet.

4. House 114 was called a household. It is a zadruga.

5. Zvonimir Vukelić was not a casualty-list death.

6. The Karlobag material was filed under the wrong family.

7. Two letters went to the wrong archive.

8. The search did not fold given names.

9. Forty Latin case-endings were blamed on a transcriber.

10. A name was called resolved on the strength of a sorting bucket.

11. Her surname is Gergacs, not Gerkacs.

12. A misreading of my own, corrected at higher resolution.

13. Five links were built without the site's base path.

14. A second wife, or a second man — the Orešković join gets worse.

15. I blamed the download for four thumbnails. It was the upload.

16. A Karlobag baptism was attached to the wrong Caetano Bacchi.

17. The Frontier colonel's name was misread as Ballinich.

18. Four sorting buckets were published as ancestors.

19. A direct ancestor was recorded as dying eight years before her wedding.

20. Five stock graphics were published as portraits of ancestors.

21. The gazetteer was counting one village as several.

And the record-level corrections

What is deliberately still wrong