Remove the unsupported judgement from the working reference list, not the raw image, and do not replace scratch with good merely because the first label was a guess

Record the physical defect locations when those decisions come back. Otherwise a comparison can remove one bright streak and everybody celebrates without checking whether the real defect is still visible somewhere else.

Quality has examined the identified faceplates. One has a small real edge defect away from the labelled streak. Another streaked plate is accepted. The working references now reflect those decisions, with the regions marked.

Can you repeat the presentations in the old images? I mean with enough holder and lighting information to reproduce them, not by trying to make a new picture look similar.

That distinction is worth keeping. If the old setup cannot be reproduced, start a controlled comparison with the retained plates and preserve the old images as limited history, not a measured baseline.

We can identify the holder but not all the old lighting conditions. New comparison will start with a documented setup; it will not claim to reconstruct the original streak movement exactly.

Include repeated removal and seating as well as a held-plate lighting comparison. They answer different questions, and changing both at once could hide which condition made the edge defect difficult to see.

And let the accepted faceplate visit the awkward views too, so the purchase decision is not based only on making the rejected example look obvious

First repeat-seating comparison produced two visible rests against the holder. I have stopped treating those as small angle variation until tooling says which rest is intended. No lighting purchase proposed from that result.

Does the holder drawing establish the intended contacts? If it does, the alternate rest can be shown against a physical reference instead of asking tooling to interpret glare in the photographs.

And keep the alternate rest as an abnormal-presentation example if that is what the review finds. Correct references alone will not stop a badly seated faceplate reaching an apparently normal image.

Drawing identifies the intended contacts. Tooling found a removable packing shim left under one support during the examples. It was not part of the intended assembly; removed under the setup procedure and its original position recorded.

Then recheck presentation with that corrected assembly before attributing anything to illumination. The physical change needs its own result, not a combined before-and-after with new camera settings.

Was the shim present in every original image, or is that still unknown? Finding it now does not date its arrival in the holder

Unknown in the oldest images. The documented comparison after removal returns to the intended contacts consistently for these retained plates, but edge visibility across the required range still needs review. I am not claiming the shim explains the whole folder.

Thanks for checking that limit. The next imaging comparison can now start from a known holder condition instead of inheriting a story the old pictures could not support.