Four bezels are retained with their image groups. We can locate the streak regions relative to mounting holes. None of the labels came from physical inspection, and no presentation notes were saved.

Can the fixture owner check how the bezel seats before you alter lighting, so the imaging comparison has an actual presentation condition to repeat?

I'd keep all four available through the comparison. On our handling work, a later tooling change once meant the supposedly same setup was not the one in the earlier photographs. The physical pieces and the fitted arrangement need identities, not just an image folder named latest.

Does one bezel repeat its own presentation, or are you only comparing different bezels?

That is a genuine presentation issue to investigate. It still doesn't classify the surface marks. Let quality and tooling answer their respective questions rather than waiting for a holder fix to decide which of those labels was true

Quality accepted three bezels and rejected one for a dent at the rim, not the bright streak. Tooling found debris under a removable holder support and cleaned and reseated it through the workshop procedure.

Did repeat seating change after that work, or have you only found and removed the debris?

The reviewed static comparison no longer shows the alternate rock on those four pieces. That is the observed holder improvement. We still need imaging to show the accepted finishes and rim dent reliably under the chosen conditions.

Keep the troublesome old presentation as a diagnostic case as well as checking the intended seat; a cleaner holder should not erase the evidence of what an incorrect seat looked like.

Does the person who changes the support know how its seating is checked now? That small maintenance detail could matter more to the next image session than another careful folder arrangement.

Tooling added the support check to the maintenance note and showed it to our changeover operator. Old images are retained with the obsolete scratch labels marked unverified; the current reference list uses quality's physical decisions.

Then the label correction and holder finding are concrete results. The remaining image test should include the accepted pieces that made strong reflections and the rejected rim, with settings and presentation recorded. Neither neat images nor a stable holder alone establishes inspection capability

Agreed. The next comparison is set up around those two kinds of error, not around making every image look similar. No threshold chosen yet, but we have stopped treating the old captions as the answer.