The old sheet only names projection and gives no individual blank identities; we have retained new identified samples for the separate length and repeat-seating comparison

Have the machinist confirm what establishes axial seating in that setup and how the reference condition is checked. You need that definition before interpreting a repeated projection change as a loading fault, and the old unpaired readings should stay unpaired.

Were all the old entries measured from the same face, Eli? It might not be recoverable, but I'd ask. Two people measuring from different convenient edges can create variation before the blank or seating even enters the argument.

Then compare repeated readings within each identified blank before comparing the different blanks. Those answer different parts of your question.

And retain any observed seating difficulty with the measurement rather than discard it as a bad attempt. The current study should explain the work needed to establish the condition, not only the readings after somebody has made everything behave.