Hello,
I did assume results were from actual measurement, not simulations..... Did you do your gain simulation for the LNA alone, or with the mixer as load?
In real measurements, attenuators can be used to show specified source or load impedance to a circuit when the actual source or load is not known, or not what you want. Off course they are also used to attenuate signals to avoid overload to measuring instruments and circuit. If you change attenuation and the quality of the attenuators is no good, you can get gain steps (like in your situation).
I don't know your circuit exactly, so I cannot say whether 772 Ohms is something to expect or not. It seems high to me (given the 2 GHz). I don't know RF Spectre, but you may experiment with accuracy settings. In your case it looks like a simulation accuracy issue. Are you sure you did correct voltage to dB conversion (in case of manual calculation)?
I hope that some Spectre experts will reply.