Post History

PCB antennas are not designed by electrical engineers, but guys with pointy hats and wands  They use trial and error after starting with the result of some mystic ritual including lots of mumbling and waving of the afore-mentioned wand.

OK, that's not completely fair, but there is still a high component of trial and error to tweak the final radiation pattern and impedance.  Simulation software can go a long way these days, but the good stuff is still quite expensive, and out of reach of many engineers that don't specialize in RF.  Not everything can be accurately modeled, so the result of a simulation is only a starting point.  Testing needs to be done with real hardware.

It is easier to model antennas as a series of straight elements.  Not all simulation programs handle curved segments, or handle them well.  Straight segments are also easier to lay down.  Sure, any competent modern PCB layout software allows curved segments, but these are all more work.  Tracks with straight segments are just easier.

You have shown some examples with angled corners, but without an analysis of how well the result works, those examples are meaningless.  How do you know the curved ones are optimal and the cornered ones not?

Note also that your main point is about impedance discontinuities, but that's quite different from "bad antenna".  What matters is the overall impedance presented to the transmitter (or receiver), the radiation pattern, and sometimes the polarization in each direction.  If an antenna achieves the desired result, why does it matter that this is done with some impedance discontinuities along the way?  You haven't provided any reason we should care about impedance discontinuities by themselves.