I took some time to put the video data into graphs, and in these graphs the numbers sure look official 
, watch out for good-looking graphs!!! In each video I turned the current up to a fairly high level and then turned the current down to minimal and started over again but now until *poof*, I did this to see if any 'burn-in' effect (lowering of the Vf at the same current) could be detected. It turned out that the second run both with the XP-G2 and the XM-L2 gave almost exactly the same numbers, so no burn-in effect here (Perhaps a burn-in effect will take place when the led is driven way over its maximum output).
Take notice that the output lumen estimates were calculated from ceiling bounce numbers that are derived a bit different with the two leds, the XM-L2 had a big 49mm reflector on top without lens, the XP-G2 had a small 20mm reflector on top with a AR-lens. The reflector size should not make a difference for the lux-reading, but the lens does a bit. (sorry I messed that up a bit in the experiment)
I believe that the voltage reading coming from the power supply in the XM-L2 graph -although a bit rough- does reasonably represent the actual Vf of the emitter at that current. The voltage in the XP-G2 graph was measured more accurately and directly at the (thick) led wires.
I just used the current reading of the power supply for the graphs, I did not check that with a DMM (should have done that). But for now I see no reason to distrust the current numbers.
Here we go:
