The two inout signals, at least an examplenof them, and the desired output should be clear, if you can grab a napkin, a lipstick and a camera and post it here would do.
I imagine a pretty known rail waveform, I imagine dc+squarewave, at 5MHz the bandwidth would probably be noticeably limited but let's say square wave. So its an elevated square, with base voltage half the value of the peaks. This is my first guess.
The second signal, the input signal let's call it a 500kHz sine, but would be whatever 500kHz BW limited signal.
Now, you want the sine to be chopped by the square, but keep the relation, not just clipped when the square is low and sine is high.
Is al that the case?
Then you need your output stage to behave as a constant resistance, variable with the sinewave input signal. The more controllable way would be to add the rail voltage in some way to the input, I'm thinking something like a conatant resistance dummy load right now, recently discussed around here. I don't know about that freq range with this loop but I guess a known load would help a lot compared to a test equipment where load could be almost anything, like an indian cow on new year...
If you want a more open loop approach and still get some kind of predictable, precise output it can get really tricky, as would probably be dependent on mosfet parameter, changing a lot from device to device, temperature and moon alignment.
JS