1) I think that method is quite OK. I think it is accurate as long as your design is more or less matched to 50-ohm. Outside of your design's bandwidth, I suspect the answer won't be right since there will be reflection. http://www.designers-guide.org/Forum/YaBB.pl?num=1233236788/1#1
2) This answer is just an IMO...I think it needs to be able to drive 50-ohm, but doesn't need to be matched to 50-ohm. In the first place, one wavelength at 30 MHz is 10m (free space). Also, I think you would only have problems if your load is unmatched, but in your case, its the analyzer (50-ohm). So for example if your load was not 50-ohm, then there would be reflection at the load, and when the reflected wave reached your "unmatched" device, it would be re-reflected. I think the concern is that if there is too much bouncing around, the resulting distortion can cause ISI.
3) For the output, just do whatever you did in the full version. For the input, if your input impedance is high, then you will get an automatic 6dB gain (it will be slightly less). However, if you add a 50-ohm matching resistor, then it will affect your NF. If it were me, I would not add any matching. Just make sure you interpret the s-parameter data properly when you want to de-embed the gain.
A lot of IMOs in my answers. Maybe somebody can give their opinion.
hope it helps,