Depth readings from a fixed face to each blank end. No individual blank lengths recorded.

Then the old list is missing one of the quantities needed to interpret it. Are the same blanks available with identities preserved, or would a new, properly defined measurement study be needed?

Who owns the loading and inspection procedure for that study, so the person collecting numbers has a consistent method and knows what to do when a sample gives an unexpected reading?

The machining and quality owners should agree it together. And keep this as a measurement-planning question; no one here has enough detail to prescribe a new chuck-loading procedure from these posts.

Check blank end condition in the study plan too, where relevant to the datum. Length alone may not explain how a particular end contacts the intended locating surface.

Yes, relevant end features belong with the sample description. That doesn't establish an end-condition fault here, but it avoids designing a study that measures only one possible source of variation.

The samples lost their row identities. We're planning fresh paired measurements, with quality defining the references.

Give each sample a label that stays with its measurements. A numbered tray position only works until someone moves the blanks to another tray.

Agreed. The old readings stay as background. They don't establish a robot problem or a seating cause.

Add that limitation beside the old sheet itself, so another person does not lift those numbers into the automation proposal as proven repeatability data.

Keep the gap visible. Ask quality how it should be excluded from the paired comparison.

And count the failed records separately, otherwise the method can look beautifully manageable after all the inconvenient rows have vanished.