And check the drawing and fixture revisions, not just the part name. My own bracket review has the same trap: a familiar description can make different inspection stages look like one continuous record.

Who put those pages together in the first place? They may know what the grouping meant

Do not rely on the folder order as chronology. I have had inspection exports sorted by filename, then read as a process sequence. Preserve the original bundle while you establish the event order.

The bundle was prepared for the proposal, not kept as a traveller. The alignment page names a fixture trial. The released sheet names a production assembly. I'm checking whether those identities connect at all.

That is a significant gap. Do not calculate change between them while the link is unknown.

You can still use the released failure as evidence that an identified assembly missed its requirement, provided its method and drawing references are sound. It just may not describe the outcome of that fixture trial.

Quality found the original index. They are different assemblies. The proposal author grouped them as examples, then the accompanying summary described them as before and after. I've asked for that summary to be corrected.

Does the correction record belong to the released failure, or is that another loose example?

That pair does match. Same assembly, failed result, correction entry, accepted result. The operation itself is still unexplained. Incoming fit-up isn't recorded on either sheet.

Then keep two findings visible: the false before-and-after link is repaired, while the corrected assembly still cannot establish uncorrected production capability. One tidy summary sentence could muddle them again.

Has the proposal author actually replaced the summary?

Yes. Revised today, with the fixture observation separated from the production assembly's correction history. The comparison that started this thread is withdrawn.

For the remaining study, let quality specify final inspection conditions and fabrication describe preparation and restraint. That gives the new cell real requirements without pretending these unrelated historical pages supplied them.

Those owners are assigned. No fixture choice until we have that basis. Thanks Rebecca; checking identities stopped us arguing about a dimensional change that the bundle never showed.

Worth keeping the withdrawn summary marked, so someone doesn't retrieve it and restart the same comparison