Hi @jp2,
Concerning your problem, I would also have the same response as you have : your problem seems to involve nested then crossed factors.
One option to see which model type sounds reasonable is to launch the Variability Chart platform, and continue the Gage R&R analysis in the two possible models: Nested then Crossed or Crossed then Nested (full crossed or full nested won't be of interest here), and look at which interactions are present in the Gauge R&R panel (see images).
You can see that interactions present in the Nested+Crossed do correspond to what you can effectively measure : Part x Operator and Location x Part (since all combinations are done for these two interaction terms).
In the Crossed+Nested model, the interaction Operator x Location is mentioned, but is in fact not fully estimable, since not every operator will face the two possible locations.
You can have a look and test these options (scripts saved) with the test dataset I created for your case.
Hope this help you,
Victor GUILLER
L'OrƩal Data & Analytics
"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)