Sorry I didn't have time to thoroughly test it, but here's what I got with one strong push.
The generator is connected to a 1N4007 full bridge, 47uF, and various resistors.
The graph is Power (W) against Load (ohm).
Power was found using V^2/R as suggested by kak111.
The power was half of what mister_rf initially proposed @ 0.06W which is true as the LEDs do light up for about half a second or so.
I couldn't test with more optimised components for the time being as the lab closed till next wednesday. I apologise for this.
I'll properly test it with a bat85 schottky bridge rectifier, and focus more on the 100ohm-300ohm region.
I'll update this thread later next week when I've got the proper measuring tools to work with.
PS: For those watching here's the data I got for now.
V I R
2 0.02 100
3.28 0.018 180
3.44 344u 10,000
3.44 73.19u 47,000
Without the bridge, the AC waveform looks like below:
PPS: How do I change the thread title? Can any mod change it to "Extracting maximum power from handheld LED flashlight?" ?