When doing Power amplifier measurement via PNA, I understand use low source power on port 1 as to not damage the B receiver.

But I don't understand why use high source power on port 2 to get

valid S12 measurement or enough power at the A receiver.

Many thanks.

But I don't understand why use high source power on port 2 to get

valid S12 measurement or enough power at the A receiver.

Many thanks.

In a most particular case, suppose S21 is high (like 70 dB) and the amplifier S12 is -90 dB, so the amplifier is stable. If you set the source power low (say -60 dBm) to measure S21 and not overdrive the B receiver, and you also had port 2 set to -60 dBm, then the A receiver will see (-90 -60 = -150 dBm) signal for the S12 measurment, but suppose with your IF BW the noise floor is -100. Then S12 will be very noise at a level of about -40 dB ( -100 -( -60) ). Now, in the math computation, there is a term to correct for raw load match that goes as LM*(S21*S12) where S21 is 70 and S12 is -40 so the term is +30 dB (where in fact, the amplifier S12*S21 is -20 dB). This will cause a huge error after correction because the raw S12 data is completely bad.