The covers are retained, but some images have only a folder date. I've asked quality to match what can be matched. I mean actual identity, not the cover that happens to look closest.

Keep the unmatched ones out of the tuning references. They can still illustrate an unresolved observation (just don't let the lesson folder promote them back to confirmed defects).

And keep the full images with capture conditions. Once quality identifies accepted and rejected samples, the presentation comparison should show whether those known conditions remain distinguishable, not only whether the troublesome reflection becomes less bright.

Thanks Will. I've separated the unverified lesson images from the proposed reference set. Quality decisions are pending, and the holder comparison hasn't started, so there's still plenty left before threshold work.

Keep the reason for that separation in the record too, so someone does not copy the old images back merely because the folder looks incomplete.