On my 3rd TN31 mod, getting about 8-9% lower output. swapped heads, issue followed the head (not the batteries or carrier or tail switch). Amps at the carrier is same as other units, so, think'n it's the emitter - will swap star/emitter with my good TN31 to isolate it down. Maybe will take a LED amp reading at that time using the solder methods. Ohh - also, I removed the stock resistors on this light. Tried both a solder only (blob), then put in copper braid for jumping the pads from the resistors - either method got same results. So, can't imagine removing the stock resistors cause this lower than expected result -- on the other lights, I left the stock resistors in and simply layed a single copper wire over each, and soldered them in.
Looking at my very unofficial LED tests (https://budgetlightforum.com/t/-/20318), the U2 1A from IOS did much better than the IOS U2 1C's, also about 8%, though tough to compare with one on alum the other on a Noctigon, but at 2A, should have been about the same I believe. Now I'm wondering if the IOS batch of U2 1C's wasn't so good? Should find out this evening...
Update: Crap! Looks like Match's tests do show an alum vs copper difference at 2A -- don't have the raw #'s, but from the graphs here: https://budgetlightforum.com/t/-/16621 I'm guestimating 6% (maybe 775 vs 825 maybe?). It's in the ballpark...