I've got our FR5 bezel samples in numbered sleeves, which is an improvement on my old loose-box approach, but the same bezel still gets called scratched or clean depending on how it sits under the light; what would you ask quality and the fixture person to check together?
Have quality identify the actual surface condition under the agreed viewing method, while the fixture person checks how the bezel seats. Then repeat the camera views with the sample identity and seating recorded. Otherwise the two people can each solve a different problem and leave you with the same bad labels.
Quality accepts the bezel; the bright line has no matching reject feature on it, and the fixture person found it can tilt against one of the side contacts, so they're reviewing that seating before we collect more images.