Now let me say I definitely enjoy the creativity and dedication of Maslownians 
In the last 2 years, there were a lot of talks about calibration, searching the set of necessary parameter, finding how to set those parameters in a simple way, yet dealing with everyone having made their own slightly or vastly customised frame.
There were posts about getting oval shaped circles, parts slightly distorted depending on where on the work space they were cut, observing wavy “straight” lines along the top edge, questions about how to deal with lower corners inaccuracies. And so much more…
I decided to take part into the “understanding MaslowCNC parameters” ground work already largely undertaken, and at the time still having some challenges ahead. So for a big while, I was focused on the list of sources of errors initiated by @dlang. My target was to understand what matters then find ways to tune accordingly a Maslow CNC until cnc’d parts would fit well together, while only caring to stay away from bottom workspace corners.
I’m getting close 
I said I tuned my Maslow up to a point I can’t measure an error with my tape measure.
But how good is that? To find out, I cut some parts meant to fit like a puzzle.
The top part mating edges were cut on the workspace at top center: Where the chain angles are low and sled subject to some vertical error. That may be why the fit has a slight gap:
The other edges were cut elsewhere. For those, I like the fit:
Details about the project:
Drawing parts with inkscape, drawing a set of connected polygons, then for each one use menu “Extension/generate from path / pattern along path”. To place the zipper pattern around each polygon. Then tune up the patterns to close the shapes.
Here is a sample inkscape file. 
Then exporting to FreeCAD, extruding to 6,35mm thickness. Setup a CAM job, define a tool, use the “face contour” operation, then add ramp and tag dressup to create paths. Simulate, then generate Gcode.
I use a 6.35mm cylindrical bit compression bit (I should’nt - i got tear out along some edges. A straight bit is said to be a better choice for plywood))
First pass is clockwise (climbing) at 600 mm/min on X/Y displacements. Each pass is sightly over 1/8 inch. Due to router bit forces resulting when routing through plain material, the clockwise part contouring direction pulls the bit slightly away from the part, making it thicker.
Second pass is a full depth counter-clockwise direction (conventional) to shave the part to accurate dimensions: While there is almost no material in the path, the router bit position error is reduced.
I’m using about 6 tabs of 4 mm length and 4 mm height to hold the pieces during cuting. That is maybe a little too fragile.
I also found that -when shaving the part- my router bit needs cutting 0,24mm more than the 6,35 mm diameter. So I set my CAM program (FreeCAD) to use a 6,1mm diameter cuting tool, but I really just use a 6,35mm one. That yields perfect zipper tongs and slightly wider notches that can press fit into each other.
(edit) To find out this 0,24mm tuning, I had to cut a few samples then find out the real tongs and notches width using a caliper. Tongs were too wide and notches were too narrow. To make sure the widest would fit the narrowest, the difference was roughly 0,24mm when using some spreadsheet math.
(edit) Let me point out that my motor spacing is 3501 mm, so chain sag effect is significantly reduced in the lower corners of the workspace. But that accuracy point only makes a difference near the lower cormers.
