The stored note of 43 mixed cells is in llm_double_coding.json.
It was saved without review identifiers.

Re-parsing the saved Yandex pages with the same splitter
("N. seviye şehir uzmanı Abone ol", then cut at reply, photo or like)
finds 42 eligible cells that still contain the string "seviye şehir uzmanı".
All 42 match a row in adjudicated_labels.csv (S13_marker_mixed_cells.csv).
The forty-third cell in the stored note is not reproduced and cannot be named.

A narrower rule, a second full dated stamp (day, month and year) in the saved cell,
marks 20 of those 42. Those 20 review identifiers are exactly Supplementary S12.
The other 22 contain the unsplit reviewer marker but not a second full date.
That is why the exclusion file has 20 records rather than 43.

Of the 42, 32 lie in the original 15-hospital set
(6 navigation-related marks and 1 walking-or-mobility mark).
Across all 566 texts the 42 carry 7 navigation-related marks and 1 walking-or-mobility mark.

Exclusion of the 20, unchanged from the previous revision:
546 texts; 15 hospitals; 472 primary texts; 30 and 26 marks.
Navigation-related rho = 0.464 (p = 0.0837; Holm p = 0.1000).
Walking-or-mobility rho = 0.520 (p = 0.0500; Holm p = 0.1000).
Does not replace H1 or H2.

Exclusion of all 42, cells dropped rather than split:
524 texts remain. Konya falls from 16 to 14 texts and Manisa from 16 to 13,
so both leave the 15-text rule.
Rank test on 13 hospitals and 427 texts (22 and 23 marks):
navigation-related rho = 0.679 (p = 0.0140; Holm p = 0.0280);
walking-or-mobility rho = 0.649 (p = 0.0196; Holm p = 0.0280).
Removing Bilkent and Etlik leaves k = 11:
navigation-related rho = 0.552 (p = 0.0807);
walking-or-mobility rho = 0.390 (p = 0.2388).
This row changes the hospital set. It does not replace H1 or H2.

Same original 15 hospitals, including the two denominators below 15:
454 texts (26 and 25 marks).
Navigation-related rho = 0.379 (p = 0.1628; Holm p = 0.1628).
Walking-or-mobility rho = 0.591 (p = 0.0233; Holm p = 0.0465).
Not the pre-specified 15-text rule. Does not replace H1 or H2.

Permutation: 100,000 shuffles, seed 22, the same procedure as S12.
Published H1/H2 remain rho = 0.512 and 0.513, Holm p = 0.1040.

Remaining limitation: the 42 cells were dropped, not separated into
reviewer-level texts, because unique Yandex identifiers were not retained.
The unreproduced forty-third cell, if it was a further mixed cell,
is still unnamed inside the published denominator.

No review wording is in these files.
