Keep acceptable surfaces in that set as well as confirmed defects. Otherwise improving the image of a scratch tells you little about how the ordinary reflective finish will behave when the part is presented again.

Record the seating arrangement too. Small presentation changes may come from the holder as well as the commanded pose, and the reference comparison needs to distinguish those conditions.

Who keeps the doubtful pieces out of the decided reference set while quality examines them?

Quality has the identified pieces now, with undecided ones separated. The images did not record whether the bezel was fully seated; I have asked the imaging team to preserve that in the new comparison.

Keep the old images, labelled with that missing condition. They explain the question but cannot supply the missing seating evidence.

Movement with presentation suggests an imaging effect, but it does not prove the surface has no defect. Quality needs to decide the physical condition independently of how cooperative the latest image looks.

Chloe, yes, I meant remove the unsupported label, not declare the bezel good; the physical decision is still quality's

When they decide, locate any rejected feature against a recognisable part feature in the images. That lets the specialist ask whether it stays visible while the highlight changes, instead of chasing whichever bright stripe is strongest.

Quality accepted two of the strong-streak examples and rejected a third bezel for a faint physical scuff at the rim. The rejected feature is now located separately from the moving stripe.

Did the seating check find actual movement in the holder, or is holder variation still a proposed explanation?

Still an explanation to check. The sample decisions are complete, but I have not received a controlled seating comparison or a revised imaging result.

Then thresholds stay untouched. The next comparison needs the accepted streak examples and the rejected rim scuff under recorded presentation conditions.

Will the rim feature be identifiable to the person repeating the capture without Bruno standing beside them?

Yes, quality marked its location on the reference record against the part geometry. Thanks, Daniel; we now have a specific faint defect to preserve, not just a request for less glare.