The sampling SE and the floor, both of them. The SE from the table, labeled as one standard error; the floor as the larger of the sampling threshold and the observed repeat difference. Reporting the observed difference alone as the floor mistakes a lucky draw for a limit.
Write the interval into the sentence, and who re-captured. "Repeat capture after 4 h 40 min by the same Team differed by 4 pp; the floor at this station is 18 pp." The observed difference means nothing without them.
Bleached coral is coral. Score it hard coral. There is no bleached class and no condition figure in the Entry.
Indeterminate is an answer. Where a frame cannot support the call, say so. Never assign it to whichever seems likelier.
The camera's file is the record. Raw if your camera can write it; otherwise the JPEG the camera wrote, untouched. Anything you edit, crop or export afterwards is a derivative. Say which you kept.
Per photo point, not per station. Your two photo points will not read the same, and that difference is information. A single station-wide figure with one frame's uncertainty attached claims more than you have.
Record a lesion, never name a disease.
Nothing above Level 3. Every statement beyond a measured value carries a level: what you measured, what a cited source says, or what a cited source plus a stated assumption suggests. This reef is degraded is a diagnosis from one visit and does not appear in your Entry.