After reading this I took some depleted Sony 18650 cells and ran two them down in my J12. These cells were depleted previously down to 3V in a capacity test, and deemed too little capacity to be any use. But they're suitable for tests like this when I want to see how a light behaves when the cells run down. The cells had rebounded back to 3.8V.
I ran the light tailstanding for a few minutes, enough to get the head very warm. Took a tailcap current measurement and the boost circuit was pulling 5.5A. I noticed the cells were VERY hot at this time. Too hot to comfortably hold in my hands, so I stopped the test here (fearing some kind of melt-down) measured the cells and they rebounded back to 3.5V each.
Paused here and let everything cool down for 20 minutes.
Ran the test again. Cells started out at 3.5V each. Re-loaded them and ran them in HI for 4-5 minutes (as I type this), long enough to get the light warmed up again. Measured current draw, and it was still pulling 5A, down a little ~.5A from before. The cells were once again very HOT, measured cell voltage and it had dropped to 3.3V. So things seem to slowly be dropping, as the cells deplete.
Paused here and let everything cool down again.
Ran the test again. Cells started at 3.35V. Only this time (knowing the cells will be completely depleted) I ran the test without the tailcap, passing current straight through my Fluke meter. Current topped of at 7.2A for a brief moment (maybe 10 seconds) and started to drop rapidly down to ~1.75-2A. Went over to the wall and confirmed HI output had dropped to MED, so the circuit was doing what its supposed to. Truncating the HI mode. The other thing I noticed was mode switching began to "flicker" a little between modes as I flipped. It wasn't a crisp HI-MED-LO cut. Pulled the cells and they measured 2.85V each, and they rebounded back to 3.25V after a 3-4 minute rest.
I Re-loaded the cells and ran the light again. At this point its clear the circuit is dropping both HI and MED output down to a lower mode, maybe a little brighter than LO. At no point did the light output cut off completely. I stopped testing here, bad/dangerous/explosive things can happen running Lithium cells depleted this far.
BLF'er benkie and a couple others have been very vocal about the existence of fake Trustfire TRJ-12 and J18 lights... and from reading others comments and reports that appears to be a very likely scenario. Personally I think it goes deeper than that... I think it has to do with the driver board supplier, and their ability (or lack thee-of) to manufacture the same DC-DC circuit form batch to batch. I have always wondered why they take the extra time to scratch off the component ID and BIN numbers from the SMT components. After reading threads like this HERE , I personally believe the driver board manufacturers are stuffing the circuit boards with whatever components they can get their hands on. They may not always be 100% compatible or identically capable. By the time you add together the cumulative sum of inter-mixed parts on each board, you end up with the variation we are seeing across the population.
Anyhow... I think this particular J12 from the aliexpress $54 batch is a good one. Just be careful with unprotected cells, and don't run them all the way down. Watch out for HOT cells, HI mode dimming down to MED, and a slight "flicker" between modes, these are all signs the cells are fully depleted.