Gain differences between HFSS simulation and test results

Hi, everyone.
I have simulation a LNA in HFSS using Rogers5880 substrate, the following fig is the test result(in blue) vs the simulation result(in green). The two have a good agreement within 13.5Ghz, but after that they are quite different, especially in the frequency of 15.28Ghz(mark 1).

Can anyone give me some advices how this happen and how to minimize the difference?

how do you manage to simulate a transistor device (LNA) in HFSS?

Sorry, I didn't express myself clearly.

I've simulated the passive part(PCB) of the LNA,not the active circuit

you are measuring on a very large bandwidth. Is your Network calibrated correctly on the whole BW.
On the other side, I would prefer to redo a simulation around 15.28Ghz with some discrete points.

Thank you for eastern.
I also did a calibration from 14Ghz~19GHz,and the results turn out to be the same with the wider band.

I'll redo a simulation as you suggested.

why thanks, if the results are the same ?