I'm questioning the labels in my FR5 faceplate set too, so I'm not treating these UR5e trim-plate streaks as established scratches. What should make the references and imaging comparison defensible before thresholds?
Keep each physical plate identified with quality's decision, then compare controlled presentation and lighting using accepted variation as well as a confirmed reject. The support needs an intended seat; several photographs of a plate rocking differently are not independent defect evidence.
I've asked quality to judge the retained plates and tooling to define the intended seat. That keeps physical acceptance separate from the changing pictures; neither the scratch labels nor new thresholds are approved.