Hi,
You are right that PXF result is voltage gain, s21 is Transducer Gain not directly Power Gain. The value of the resistor of the port should be the input impedance of your next stage to make a measurement of Voltage Gain.
Hope it helps,

Hi manfred,
1.This will be quite teoric since it comes all from microwave theory and how spectreRF implements it:
In general you have several "Power Gains"
Transducer Power Gain(unilateral-->s12=0):
G_TU=((1-abs(Γs)^2)/abs(1-s11*Γs)^2)*abs(s21)^2**((1-abs(Γl)^2)/abs(1-Γout*Γl)^2)
Where:
Γs:source reflection coefficient,Γl=load reflection coefficient,Γout=output reflection coefficient.
Since spectreRF uses as reference impedance of each port the port impedance Γs=Γl=0 and G_TU=abs(s21)^2,this gain is the ratio between power delivered to the load and power AVAILABLE from the source(input impedance conjugated matched to the source, if the source impedance it's 50 ohms and you matched to 50 ohms the it makes sense as power gain).
There are several definitions of power gain a good book that has all the definitions it's "Microwave Engineering" of David M. Pozar, Wiley Editorial.
2)Yap you are completely right, if you put a port impedance of the same order of magnitude you are loading the mixer.
Hope it helps,