I’ll provide more information when I have time. The cut I made today looks dimensionally correct, just scaled down slightly. Pockets which were supposed to be 19.1 x 81.4 mm came out as (on average) about 17.6 x 80 mm. The scaling does not appear to be uniform.
I drew the file in Illustrator and output the gcode in Easel. The dimensions are correct in Easel.
I have a similar result. When I get my computer (to get cutting again) I will be updating sled including adding what I am calling a YAZAM! (Yet Another Z-AXIS Mod). Then I plan to see if, with math, I can figure out the sizes to make the SVGs so that they come out the ACTUAL size I’m going for. My idea that may or may not be able to help me with getting around advanced calibrations that I don’t yet understand. 
We’ll see!
The problem is that the distortion is different in different areas of the
worksheet.
your approach is what our first calibration attempt did, cut a square, measure
how far off it was, tweak a parameter, repeat until it is square.
then we found that as you moved around, we still weren’t cutting right.
Somewhere in GC there is a simulator that lets you say that various measurements
are incorrect (chain length, motor spacing, rotation radius, etc) and see how it
throws off the positioning. play around with that and you will see how different
errors produce different distortion, and how the distortion is different in
different areas of the workpiece.
this is why the current calibration approaches are taking measurements in
different areas (we try to get measurements that maximize the different sorts of
error) and then tries to figure out what the errors probably are and tweak the
parameters to correct them
David Lang
are you remembering to allow for the size of the bit?
if you drw in illustrator and then cut on the lines that you draw, the part will
be undersized by the diameter of the bit (technically the cut is short by the
radius of the radius of the bit on both sides)
David Lang
Are there calibration procedures apart from what’s in GC?
Yes, there is the holey-triangular-calibration that has a couple of great results reported back. You will run a fork of FW/GG that will hopefully be merged into the main branch soon. I am testing it more the next 2 days, so ask if you are taking the same path.
Easel allows you to specify whether to cut on the path or inside/outside the path. All my cuts were set correctly.
I just finished installing a counterweight tensioning system on my frame but didn’t remove the chains from the motors. I did run the five cuts calibration from GC, but skipped all parts before it since (I thought) nothing related to those steps had changed. Do previous calibrations carry over if only one setting is changed?