Two bezels are retained, each with its image group; the labels came from our own image discussion, and neither physical surface has an acceptance decision attached.

Can you locate the same physical region across the views? I would preserve that relationship as well as the bezel identity. Otherwise the discussion may compare two different regions while treating the brightest patch as if it were the feature being inspected.

Who has the retained pieces now? The image folder can't substitute for that handover.

For the imaging comparison, record the actual presentation and camera and lighting conditions. Change one variable at a time where practical. That helps explain why an appearance changes without turning the image experiment into a new physical acceptance rule.

Naomi, yes. Include an agreed acceptable example that produces a bright band if one is available. The comparison should expose false scratch responses, not only show that a known reject can look bad.

Quality found a narrow tool mark on one bezel, outside the travelling band, and accepted the other under our cosmetic requirement. The accepted bezel still produces a strong band in one recorded presentation.

Does the tool mark remain visible in all those presentations? That tells you whether the current image conditions hide a required defect, even if the physical reference decision is now clear.

No, the mark is obscured in one view; that failure and the accepted bezel's bright band are both retained as test cases, so the imaging comparison has two specific problems to solve.

And the old scratch captions are out of the current reference set?

Yes, replaced with the linked quality decisions; originals remain in the diagnostic history, while threshold tuning waits for the presentation and visibility comparison.