The first time after triangular calibration it reported 2+% error.
.
I manually entered the calculated values into settings and ran it again. It reported 3+% the second time.

Is the error supposed to be getting closer to zero with each calibration?

FWIW, I’m pretty certain the main measurements are in Imperial. If your measurements don’t converge upon the “perfect” measurements, you may need to reset the “basic settings” (manually enter the motor height, etc. via Quick Configure in Web Control). If these settings were wrong, calibration will just thrash back and forth between incorrect extremes (I ran it several times with ~15% error each time before figuring out something was wrong).

I found that M1, M2, M3, and M4 (horizontal) all converged on 38" exactly.
M5, M6, M7 (vertical) converged on 28" exactly.
M8-11 (diagonal) can be computed at 47.2" (pythagorean theorem matches reality).
M12 should be exactly 10" from the top of the stock.

My tape measure is in imperial (sigh) so my accuracy is only to the 1/16th of an inch, whereas a mm is about 1/24th of an inch. With this constraint I can get a little underneath 1% error, but really the accuracy of the measurements is the limiting factor.

The coordinates for the holes are based upon the size of the workspace found in settings (defaults to 2438.4 mm x 1219.2 mm … 8-feet x 4-feet), but they are positioned 254 mm (10-inches) from the edges regardless of workspace size.

This is the formula used (using millimeters):

x = (workspaceWidth / 2.0 - 254.0)
y = (workspaceHeight / 2.0 - 254.0)

So top left hole is (-x, +y), bottom right is (+x, -y), and so on. Basic trig can get you the measurements, but I think you figured that out… Someone with a different sized machine will get different measurements.

Interesting. I am trying to pack my plywood sheets as much as possible. The cost of both acquiring and disposing of wasted wood is annoying to me.

As a general benchmark, I’ve found that 5%-10% error margin makes 6-12" on the sides unusable. Right now, around 1% error, I can cut within 4" of the sides with less than 1/8" of horizontal drift between the top and bottom of the sheet (e.g., a straight vertical cut 4" from the left side would be about 1/8" different on top from bottom).

there is also the question of how accurate the tape is (class 1, class 2, or
worse than that), see In search of accurate measurements for a
vernier adapter I designed to work with metric tape measures (and a discussion
on tape measure accuracy). Rockler has a good imerial/metric tape measure that’s
a class 2 tape that’s not that expensive ($10)