Runtime estimate instead of meter? If you can measure the L3 you can measure the ZL. My 68 is only an average sample with NIMHs - which is why it normally uses a L-ion. I popped an Eneloop from my spare tray just to see what it said. Others have posted 68s pulling 2A on NIMH, as have I (previously with a different sample using my favorite Duraloops) - perhaps a better cell and more efficient emitter would do even better …
Ah, the ZL driver isn’t 100% efficient. Let’s allow for efficiency differential at 15-20% (and double the lux vs the ZL). Not enough to cross a 1 log human sight differential. However, lux is a difference you can see - or not if it doesn’t throw.
I haven’t bothered to try my Nichia L3 draw; no tailcap make it a little harder to measure while including body resistance. Nice deflection, though.