A scratch label that follows the lamp

Quality called it doubtful under their bench light. The image folder nevertheless labels it rejected.

Then pull it out of the definite rejects. Doubtful is a perfectly useful category while you sort this. Turning uncertainty into a confident filename won't help the lighting work.

Include an agreed good plate and an agreed defective one in that comparison. Otherwise you'll learn how your uncertain plate photographs and still have no acceptance reference.

Who will keep those physical references available after commissioning, when somebody replaces a light or bracket?

Quality ought to be involved, but don't assume they have storage sorted. We lost our reference samples in an enthusiastic bench tidy. Photos survived. The useful bits didn't.

Bench comparison done. The bright band crosses different surface features as presentation changes. Labels now marked uncertain.

That supports reflection. It does not establish that the plate has no defect.

Correct. I meant the image labels are uncertain, not that the plate passed.

Fair enough. That distinction tends to disappear when somebody copies the summary.

Have you got the two agreed reference plates yet? I'd rather see those results before anyone starts selling a threshold as the answer.

And did anyone take ownership of storing them?

Quality supplied two accepted and two rejected references, with reasons. They own storage. Imaging comparison remains underway.

That gives you a much better starting set. Include the uncertain plate as a separate investigation example so it doesn't quietly get promoted back into the reject folder next time someone combines folders.