Are you testing this on the Qorvo eval board or your own PCB design?
Some points to take into account:
- PCB and connector losses - Qorvo’s s-pars and results are referenced to the RF pins of the device.
- Temperature - Gain will drop at increased temperature. Running CW the device temperature will be higher than running pulsed as per the results in the datasheet. Read the post and reply at link below for more on this subject.
- Part to part variation. For multi stage GaN MMICs there can be a large variation in small signal gain, typically it can be +/-4dB. The amount of gain variation reduces when you get into compression at large signal.
- The bias point Idq will effect small signal gain. Worth checking the Idq is set correctly and has not varied during testing as the device warms up.
There is a useful post with more info on this subject here.