I'm looking into this now. Today I measured about 7 amps between the carrier and driver for 2 resistors, and getting about 390 kcd measured. So, I'm not getting much more. I know my kcd #'s are low, so maybe it's really 410-420 kcd, still 60 kcd (12.5%) behind. I copper braided the + spring and saw no gain at all. So only thing I can conclude is the LED-star and positioning. vinh says he's doing something about it, just no clue what exactly. I left everything pretty much stock - same star, same emitter, but re-seated the star with AS5 and lapped the surfaces, went with 20 gauge LED wires.
So if it's only 12.5% addition he's getting, in theory you could make that up in simply re-positioning the LED - I've seen those kind of improvements before, but really doubted it for the TN31, but who knows - maybe the factory/design was that messed up?? Thinking if you raise the LED, lower the profile of the plastic alignment piece, you may get a nice bump, so could be what he's doing.
He mods a ton of lights I'm assuming for profit, so he's either too busy or too protective to explain details of this mod, dunno... he's a pretty good guy and probably just too busy, but he doesn't really participate on BLF, so maybe he'd post details on CPF eventually. Now if I could take a look at Brad's mod'ed TN31 or any other vinh 480 kcd TN31 or his super K40's, it would probably be pretty obvious. He only fairly recently did these high end TN31 mods, before, he was not doing any resistor mod at all, just de-doming, wires, and a SinkPAD he claimed...
Edit: Just posted a detailed question to vinh on his help line forum on CPF. May take a while but see if I get an answer.
Ohh, also, 480 kcd is what was measured by someone on this level TN31 mod, but of course there could be a wide variance from unit to unit, so 500 kcd is in the realm of possibility for what you are seeing.