I rotated your center image and converted to gray, adjusted contrast/brightness so that my program can detect the lines properly and did analysis. (You can only trust the direction and amount, not the numbers as I don't have necessary infos like your 1:1 image,focal length nor sensor pixel size.)
It shows Red side is tilted to right and Blue side is off center to left. Please refer this with rotated image and your original to find corresponding patterns of yours.
You can actually visually see middle spikes are not in the exact middle in your corner images. This is 'part' of the problem of your corner stars and you can improve it a bit. But I think one of the main reason you're not satisfied on corner stars are from clippings by FR.
I attached corner crop captured by full frame DSLR when I tried F/4 FR/FF stacking. (You can find full image from the link I put on previous post) You can see blue circled star is reasonably round but red circled ones are not round with flat side facing to outer edges. If your corner stars has clipping enough to missing spikes, I guess your star shape will be similar to this at the corners. Interesting thing is, stock EDGE optics and their plain 0.7x FR shows that much of clipping. I may have to try 0.63x FR/FF tonight to see how much clipping is happening on my conventional SCT.
ps. My program has too many flaws and precarious to distribute to anybody. Working on it with one volunteer as alpha tester but we hit immediate road block from ASI ASCOM driver is not responding to my expectation where I don't have any ASI cameras. Dang.
Did you ever end up finding testers? If not, I have a bunch of ZWO cameras here, and would be primarily using the mask with the 6200mm Pro, once I print out the mask. Scope is an SCA260.
If you're no longer working on it, did you put the source anywhere like Github? Your app looked like it had real promise. Maybe, with some help, it could be turned into a NINA plugin and use all the camera driver integration already present in that?