My own quick runtime tests showed almost identical runtimes between a ZL14500 and a Duraloop on the two highest modes. Li-ion is easier to top off, noticeably lighter and has the 500 lumen burst. Also unlike other lights compatible with AA, you get proper low-battery stepdown, battery checking ability and all low modes with the Li-ion in the SC52.
On my only other light that I can test identical modes of Li-ion vs NiMH (my Quark AA) the 14500 would win on the high for runtime. It lost on max but not surprising considering that it has 80% more output in this mode, high is the fair comparison since the outputs are matched. This says to me that the ZLs AA circuit is better than most (which I think alot of people already knew). This light would not step-down though as the 14500 depleted, therefore it was full output and a sudden lights out.
Bottomline: ZLs AA performance is awesome but 14500 matches it and provides some perks.
I've mentioned this in the other thread, but from what I can tell, the SC52's battery meter seems to be geared for ICR Li-ions. It will indicate weak battery and automatically step the light down at ~3.7V resting - which, for a typical ICR is 80-90% depleted, and a good cut-off level.
The problem is that the light may not work well with a broader range of Li-ions since some ICRs (eg, my Eagletac's) and other Li-ion chemistries (IMR, IFR) may only be half depleted (or still fully charged) at 3.7V when the light starts to automatically step-down... ie, you lose high way too early.
Still waiting confirmation from other users though, only one other person has indicated his concern with this, but with no testing behind it.
(EDIT... just noticed, your step-down function is not working? ie, so you tripped the battery protection circuit when it went lights out?)
As far as ZL having a better circuit, I don't buy it either...or at least anything near what the difference in manufacturer specs would suggest.
I have the Quark AAX, and Eagletac D25A clicky (which I consider to be the SC52s closest competitors), and both with XML emitters, they are spec'd pretty close to each other on Eneloops. I have tested them with a DSLR lumen meter and have also run random runtime tests on the lower modes and found them to be pretty close to each other and true enough to their manufacturer specs.
Comparing the SC52, first off, the ZL lumen scale is at least 30% off the other two companies (Thrunite too) and Selfbuilt's test of the Quark AA2X, D25 series (Neutron 1A too) confirm this as well - don't know who's too conservative and who's too liberal, but of the 4 moonlight mode/AA companies I collect from, ZL stands alone.
Even considering the 30% scale difference, which is essentially imperceptible to the naked-eye, I was really looking forward to getting two to three times the QAAX runtimes on moonlight and low - the spec'd runtime difference between the manufacturers - on what are, by far, my most often used modes. Instead what I got was a one-half to one-third the measured lumen levels, on what should be the same levels as stated by the manufacturers. So, in terms of lumen-hours, both are equally efficient, with the ET right there as well, with almost the same lumen levels.
Even by ZL's own liberal scale, all my L modes are off by a mile, so perhaps I have a defective unit. That would make my defect rate 2 for 2 within ZL lights and 3 for 25+ for all my purchases.
It will be interesting to see the SC52 output/runtime graphs when Selfbuilt finishes his review of the light. Too bad he only tests the higher modes, and will not have a QAAX, or D25A clicky in his tests. The Neutron 1A should be though.