Shoulder at Power Amplifier

Hi guys

i saw something that not understandable to me
i tested some PA and under different output power levels and saw something strange with shoulders
According to the test shoulders are improving with power rising.

Like i know shoulder should be better at lower power than at higher...

please advice

It is possible to behave like this when the output match is optimized for best linearity at high output power (e.g. load-pull can be done for linearity, output power, or efficiency).

I guess it depends on what you mean by "shoulders". Could you be more specific on what you are measuring this on, what you see exactly, and what the drive signal is (constant envelope, QAM, etc)? Do you know anything about how the amplifier is biased up (class A, class C, etc)?