After I made this post, I thought of a few items. Most satellite upconvertors I worked with had 75 ohm input impedance and 50 ohm outputs. All of them had input attenuators.
The last amplifier in the upconverter is going to set the noise floor on the output.
I know these are old, but HP published a wholes series on using spectrum analyzers. The theory still holds today and they are worth a read.
https://www.hpmemoryproject.org/technics/bench/8568/bench_spectrum_docs.htmWhen I said to choose the proper resolution bandwidth on the analyzer, its going to have to be below the noise floor of the output. prudence is minimum 6 db, more than 10 db is excellent. With that said, the input carrier will also have to be clean, you will have to work the gain back thru the converter and make sure the test signal is cleaner than the input amplifiers output.
The point is, this is not a trivial measurement. It will also require a clean low phase noise signal generator. It may be easier to use a data modulator/modem, were you can kill the modulation. We use to do this by pulling our Forward Error Correction TX board and then the modem went CW on the output.
I do not have access to some records I have but somewhere in my head, I thonk the C/N at EIRP had to be clean like better than 65 db. The idea was not to have any intermodulation being thrown to the satellite.