Experimenting is always a good idea. A patch antenna with a single feed that is perfectly circular can't produce circular polarisation. There needs to be something else. I settled on the dual resonance method. The original POTY (it was not called that, that was Remco's idea) used an ellipse but that's hard to make, so I tried an approximation to an ellipse. A circular patch with notches, circular with tabs and the square with cut outs. The square was good enough and is easiest to make with simple tools, so that's what I ended up with. Other iterations had directors, choke rings and so on. Another way to do this dual resonance trick is with a feed and a tuning capacitor, but I found that really hard to get just right. G3RUH managed it but not with the waveguide. Achim did manage it but I don't know exactly how.
What wasn't obvious early on is were the simulations accurate? I was taking huge liberties with the meshing in the simulation and with the full version of CST we could probably design a more forgiving, broader band patch. In the end, its only when something is reproduced in hardware that you can be sure it works, or not. So, keep on cutting and trying because if it works in hardware, even without simulation, it works.
it was me who brough mike to the idea of the square with the cut corners (i have no lathe so peters feed was not easy to reproduce on the kitchen table) .. i calculated in sonnet lite and mike calculated in cst studio ... and end result was near identical
rest is history
difficult to build and tune ?? yes a bit ... but when exactly made shows a good circular pattern and is near spot on