Why does our accepted UR5e bezel look worse than the reject?

Just about. It is near the rim, much fainter than the band across the accepted bezel. Both have their part numbers beside the full images.

Check the rim is visible in the intended seat. A camera cannot inspect the bit behind the holder, however enthusiastic the threshold adjustment gets.

Yes, LeoBaker. I meant find the scratch in the normal view, not keep tilting until it finally shines.

Rim is visible, Zara. The largest tilt in our folder was deliberate, and I've separated it. The remaining pictures use the intended seat but the band still moves a little.

Then leave the holder out of the blame for now. Ask vision to compare lighting using both parts and those normal reseats.

Vision is doing that with the existing lamp first. They want to try another angle before suggesting a new light.

Will you keep the faint scratch visible through the comparison, LeoAdams, because we improved our accepted part's picture once and lost the thing we actually needed to reject?

Yes, same rejected bezel in every candidate. Thank you, Mia, I nearly sent only the troublesome accepted one.

A successful view needs the accepted finish range too. Your strong-band example is useful, but one accepted bezel may not represent all normal appearances.

Would quality have another accepted finish example available? No need to manufacture a huge image set before that simple comparison.

And no polishing the samples to make the camera happy. Keep their production finish and agreed preparation.

Quality added a duller accepted bezel. The first lamp candidate hid the real scratch as Mia warned; the second keeps it visible across the repeated seats so far. Vision is checking that candidate further.

Keep the failed first candidate with its reason. Someone revisiting the bright-band problem could otherwise select that attractive image again.