Repeat reseating of the same identified blank too, because different blanks alone cannot show whether one blank returns to the same position.

We have the manual study agreed with the setter now, including same-blank repeats and the setup identity; no automatic trial hidden inside it

Check the measurement repeatability as well. I mean repeat the reading without moving the blank, then compare that with the reseated readings. Otherwise a wobbly measurement method can look like a wobbly seat, and you will be chasing the wrong thing.

Keep those two repeat types in separate columns. Also note which end or feature defines blank length if the end faces are not equivalent.

Who will judge whether the reading differences exceed the measurement method's uncertainty, rather than just looking different in the spreadsheet?

Quality is doing that with the setter, and Aaron, thanks for separating repeat readings from repeat loading because my first draft called both of them repeats

First study shows blank length explains part of the projection spread, but one identified blank also changes projection after reseating beyond the reading variation quality observed; I cannot call it stock length alone

Keep that blank for the setup review. Does its end condition differ from the others, or does the same seating change occur with more than one blank? Don't jump from one useful finding to a stop adjustment.

And retain the unchanged setup record with the results. The next inspection should investigate that seating difference before anyone uses an offset to make the projection column look uniform.