We have three faceplates retained, with images linked to each, but the labels were assigned from photographs. Quality has not examined them. One plate appears in both labelled groups depending on the view.

I wouldn't relabel that one accepted just because the streak moves, Daniel. It may still have a real defect elsewhere. Keep the uncertainty attached to the physical piece rather than replacing one confident guess with another.

Has that contradictory example reached the training folder? I would mark its status there as well; changing the working folder may leave the old lesson intact.

Caleb, we had that exact problem with a copied slide deck. The updated references were fine, but the lesson still taught the original error with a large red circle around it.

Did you replace the slide or keep it as an example of uncertainty? I can see value in either, provided its purpose is unmistakable.

Caleb, one training draft uses it; I've marked it unverified and linked the original views. Emma, I haven't assigned a new verdict. Quality will judge the actual plates under the appearance method.

We kept the old slide separately as a lesson in why images need references, Caleb. It was removed from the acceptance examples. The red circle had become quite persuasive without deserving it.

Once quality has decided, compare presentation and lighting separately around those identified parts. Include the observed awkward seat if there is one; don't assume a new lamp is the answer because the symptom is bright.

And retain the real defect visibility as the aim of that comparison. I have seen an improved image mean only fewer reflections, while the faint mark that mattered became harder to see.

Quality has accepted two plates and rejected one with a physical scratch. The rejected mark is now located against its image. The holder and illumination comparison has not begun, so there is no threshold decision yet.

That gives the lesson a usable reference, once the imaging work is ready. Keep which physical decision each example represents beside it, not only in the filename.

Does the selected reject include the faint scratch range your requirement covers, Daniel, or just an obvious mark? That is still worth establishing before calling the reference set representative.