I've more or less come to the same conclusion and have built a procedure to
print a bunch of measurements. When I have a close one I'm homing in on the
closest I can get. I suspect that it alters with printers as well.

This is coming from trying to print a "mask" for the nut of an electric
guitar... Hence the need to go down so small. The thinnest string I've come
across is 8 thou.

When I was researching this I was suprised to see that there doesn't
actually appear to be a single exact measurement for a point. In fact there
are quite a few!