Wow.. I never realized it was driven that hard! It's actually been a light I've considered for a while now but I always end up turning down the Yezl products because many of them don't seem to get favorable reviews for build quality.
But yeah, that's about what I'm looking for in pretty much any XM-L. For me it might as well be driven to its max and 3.5a is pretty close to where it starts leveling off, based on reports compiled by Match. It seems 2.8a would be the most efficient max since that's where it begins requiring more amps per lumen.
I was exactly in a sort of detailed discussion with Match and his scientific testing does indicate levelling off at 3A, some mods done by others do indicate otherwise by both photos as well as lux meters. Lets not take those crazy mods with a full copper block and sleeve down the whole tube, this is a simple enough one : http://www.candlepowerforums.com/vb/showthread.php?312241-Loongsun-LX-9012-XM-L-Monster-thrower-(now-with-lux-readings) This is a 2-cell Made in China production light, so the voltage is 8 volts or so, there still would be voltage drop in the DMM affecting accuracy, but i mean lets take it as it is, relatively speaking for both lux as well current. The lux meter does not even have to be precise, just need to show the difference.
I had played with dry-ice cooling at a time just for fun (I probably got close to 4000 emitter lumens from 3 XM-L even with a production light and just normal thermal adhesive on MBPCB), but well of course that does not really fit into the picture in reality. But yes, my testing shows that with enough cooling and if possibile elimination of the MBPCB, more juice can be milked past 3A and the levelling off just pushed upwards the current graph.