If you had the 123x2 or the AAx2, you could look at the specs for the quark turbo on 4sevens website, which should be more or less the same for the comparable Ti models. Same emitter, same drivers, same batteries. Lux/throw would be different, but that's not mentioned on the website anyway.
Some clever math would give you an idea what your Ti 123 is doing.
Ti123 : TurboAAx2 :: regular123 : regularAAx2
(since they all have the same driver)
4sevens used the same runtime from xp-e for the xp-g.
According to Cree xp-e datasheet and Jtr1962's results of xp-g r5 (of only one or few samples), from current of 100mA-700mA, xp-g Vf is ~0.2V lower than xp-e and this gap widens at higher currents. Therefore, we should expect some increase in runtime of Quark models with boost circuits (0.9-4.2V) but it'd be hard to say about buck circuits without actual test results.
I received 2 Ti Quark 123-2 and did some runtime tests:
-Ti 123-2:
Test 1: 7h on Hi (250mA to LED)
Battery: AW17670 rated capacity 1600mAh (less than 3 months old)
Test 2: ~55h on Low (10mA to LED)
Battery: 1x Trustfire 16340 + 1 dummy cell (battery 1.5-2yrs old, actual capacity ~500-650mAh).
I guesstimate the runtime with AW17670 cell on Low will be AT LEAST twice of this, or ~5days.
-Ti 123-2 Tactical:
Test1: 15.5h on Mid (50mA to LED)
Battery: 1x Trustfire 16340 + 1 dummy cell (battery 1.5-2yrs old, actual capacity ~500-650mAh, not same cell tested in test 2 above).
I guesstimate the runtime with AW17670 cell on Mid will be AT LEAST ~31h
All tests stopped when batteries have without-load voltage of 3.5V (or slightly lower if I miss it). Typical xp-g R5 will have Vf of <3V @250mA so I'm pretty sure the lights had flat regulations during my tests.
I don't plan to do runtime on Max as the lights get hot quickly, don't want to ruin this limited edition
Note that after the 2 extended tests, one of the Trustfire 16340 cells could hold little charge and had very high self-discharge rate
so I threw it away (I expected this to happen and that's why I didn't use AW cells).