Yeah, it turns out the 500mW Mits P73 was a tricky diode to test, because it turns out it's VERY temperature sensitive. Way more than the other diodes. Even though I was using a large copper heatsink for the tests, the slightest warming or cooling would noticeably affect the measured output power. I was getting different results for the G9 if I measured it first (with the diode/sink completely cool), versus if I did it last, after all of the others.
Even though the heatsink wasn't even warm to the touch, the few degrees difference did seem to make an impact on the results, which probably explains the difference you're seeing (as the G9 curve is probably slightly on the low side). I was seeing over 1400mW with the G9 if my heatsink was slightly cool (just from sitting in a cold room), but I didn't use that data, as I felt it was a bit biased.
The better thing to do would be to use a TEC on the heatsink, to try to keep the diodes at a more stable temp for these tests, but I didn't have one handy, and I'm not trying to go nuts here. (I was more concerned with trying not to blow up diodes, as I already killed one S06J when I tried to push it to 700mA, and it popped.)
Regarding testing the new 90/120mW 520 diodes, I'll probably pick one up once DTR starts stocking them, so I'll add that data eventually.
-Andy