Help: problem about input match of the LNA between the large and low input signal

I am designing a narrow band LNA, the range of the input frequency is from 1G to 1.05GHz. And I have used SP in spectre to simulate the input match. The S11 has been reached -10dBm from 1G to 1.05GHz, wihch means the input match is ok(in my mind). After this, I used PSS in spectre to simulate the P1dB, and I found someting strange, the power of the input net ( the net after input port and before input match) is the same as the input power in the input port when it is less than -20dBm, but when the input power is higher than -20dBm, the power of the input net does not follow the power of the input port. The input match is L type: L-series and P-parallel.

My question is as follows:
1. Does this mean the input has not been matched when the power of the input is high?
2. If the input has not been matched, does this mean the result of SP is not correct and what should I do?
3. Should I use psp to simulate the input match?
4. What I should to use when simulate input match and p1dB?

Please help me , thanks!

I am not used to simulation but I know the real world. A LNA is primarily designed for a low noise operation. Its P-1 dB typically is low and it can be easily overdriven. Using -20 dBm is a typical level for DETECTORS, not LNAs. So do not do it if you can.
If you overdrive a LNA by -20 dBm, no wonder its input is mismatched- its input rectifies the extremely high signal and there is no way to correct it.
There are some LNA designs where the active device uses a high DC current for a higher P-1 dB.
Usually, receivers that should handle in put levels from -100 to -20 dBm must use AGC. Input match is important for the low noise operation. If you still need a good input match at a high level, you can use an isolator which, however, adds up some noise.

Thanks jiripolivka for reply.
However the reply does little help for me. What I want to know is that why the result of the PSS does not match the result of the SP or PSP. The result of the SP shows that the input is matched when the power of the input ranges from -40dBm to 0 dBm, while the result of the PSS shows that the input is not matched when the power of the input reaches -20dBm.

This is why I do not trust simulations. Make the LNA and test it! The real world is real, simulations are often not good.

How to define "input power" ? It's unusual...
Power is defined with impedance so you don't know input impedance hence you can not see "input power" since you don't know impedance.
PSS doesn't give you any reflection information, how do you consider that ?I think you make a serious mistake..

To BigBoss: The input impendence in my design is 50 ohms. Thanks.
To jiripolivka: The simulation is the guide of the circuit design, and I think we cannot skip it when we design a shematic. Usually, we should adjust the parameter of the shematic according to the result of the simultaion. You are very kindly to reply me, thanks.
Is there anbody who have designed LNA or Mixer before in this board to give me some suggestion? Thanks.

Which is the IIP1db of your LNA? You should try to stay far by 10 db to make sure is linear regime. In that area, your s-parameters should be valid.

To Itrlspree: IIP1db from the result of the PSS is -19dBm. In your reply, I get that S parameter may be not valid when the input power higher than (IIP1db-10dBm). And my question is that if I use PSP to simulate the S parameter, can I trust the result from the PSP, when the input power hiher than (IIP1db-10dBm)?