If you conjugate match the output you will get maximum gain. Usually in a PA you would like to get maximum output power, which says that you would like to find the load where you have the highest output power compression point. Those two point are not the same.

Yes, the optimum "load" impedance for highest output power compression is generally not a conjugate match.

Steve Cripps has a IEEE MTTS conference paper that does a nice job of explaining matching a transistor output for maximum saturated output power.

Cripps, S.C., "A Theory for the Prediction of GaAs FET Load-Pull Power Contours," Microwave Symposium Digest, 1983 IEEE MTT-S International , vol., no., pp.221,223, May 31 1983-June 3 1983

http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1130864&isnumber=250...

Hi,

of course you can use a load-pull analysis and find e.g. the optimum impedances for max. gain, max. pout or best PAE or linearity - they all will be different! So a compromize is needed.
However, I would say this is a good starting point only. Usually at some point an optimizer working on all specs is the best approach! In RF design chaging e.g. a W of the output transistor or the bias point (and bias impedance!) can change all measures simultanously. A human can work well on few parameters and specs only - a good optimizer on quite many!

Bye Stephan