OK, I've just attempted to run the procedure on the SigHound and it's rather difficult. For a start, I cannot make a normal sweep and a TG sweep share the screen, but there is a further problem - selecting a TG sweep seems to invoke a whole new setup of the instrument and the noise levels are simply not comparable. Maybe I'm missing something, but it was more trouble than it was worth! I later set the SH to do a normal (non TG) sweep and used the TG in CW mode at spot frequencies, like a signal generator. The leakage level was so low that a 30KHz RBW would not see it. I had to get down to 10KHz or less to lower the noise floor enough to see any leakage at all (using 2dB per div). At this level (around -110 - 120dBm), the analyser's own internal spurii are showing up anyway, so I stopped there.
This test was done with the units in their normal configuration, side-by-side in an instrument case but with the SMA extension leads disconnected. All interesting stuff for a Sunday afternoon!