These have a DC output, not RF so input impedance of the test equipment doesn't matter. They typically feed the vertical a scope or DC? voltmeter .
The input VSWR of the 8473C is quite good up to about 18GHz, but above this a decent attenuator will have a lower VSWR.? So the attenuator acts to isolate the less than perfect VSWR of the sig gen from the less than perfect VSWR of the 8473C detector.
However, a lot depends on what frequency range you want to operate up to.
?
The source impedance of a typical sig gen will improve at lower power levels when using its internal attenuator so the sig gen might have a fairly good source match when delivering -10dBm. I've not measured the 8340B so I can't comment on this.
?
However, it is unlikely to be as good as a decent 3.5mm 10dB attenuator that is rated to 26.5GHz. Every little bit helps when trying to minimise mismatch uncertainty...
?
If you ultimately want to use the 8473C for external levelling of the sig gen then maybe consider using a 11667B splitter to provide the feedback path as this will have very low source VSWR when the external levelling is in closed loop.
Michael A. Terrell