Sweep a batch of photo captions before publishing

Answer in [ANSWER LANGUAGE], even if the material below is in another language.

Below are photograph captions written by different people. Every caption is numbered. I need to know which ones make a claim we cannot support yet.

Rules:
1. A caption that describes what is visible is a description. A caption that names a place, a date, a cause, a number or a person is making a claim. Treat the two differently.
2. Do not decide whether a claim is true. Decide what evidence would be needed.
3. Never invent a detail, a place or a date. Never add a caption I did not paste.
4. Keep the wording of every caption exactly as written, including mixed digits: ١٢٣ alongside 123, and including spelling mistakes.
5. Every caption gets exactly one decision: publish, hold, or rewrite.
6. You cannot see the photographs. Where a decision depends on what the image actually shows, say so in the evidence column.

Give me, in this order:
A. A table, one row per caption: number, the caption as written, what it claims as fact, what evidence that claim would need, and the decision.
B. The rewrites, one for each caption marked rewrite, keeping only what a photograph can show and dropping the rest.
C. Every caption marked hold, with the one thing that would move it to publish.
D. Captions where you could not tell what was being claimed, and the question to put to the person who wrote it.

If you are reading on a phone, replace the table with one block per row: a heading line, then one short labelled line per column.

The captions:
[PASTE THE CAPTIONS, NUMBERED]