I didn't previously notice the schematic with single ended output transistors in post #6. If nothing is said about the topology and RF PA isn't explicitely stated, I associate a H bridge with class-D amplifier.
The single ended topology has a larger tolerance for different controlling waveforms, it must not necessarily be exactly inverted. Nevertheless, a PWM square wave can be much more easily inverted than delaying it by exactly a half period. Or which delay method did you consider?
So the answer about suitable way to generate the control waveform is obvious in practical sense.