The specs shown in the Sofirn VDO above read 1000 lm for 2.5 h, presumably regulated. To see if it’s feasible, I looked @koef3 SFT25R test and assumed that the OTF lumens are some 80% of what the naked LED puts out.
With 4 LEDs, one needs some 320 lm out of each to get 1000 lm OTF. That translates to some 2 W delivered to each emitter, or ~8 W total. Assuming the driver efficiency to be a likely pessimistic 80%, that should require not more than 10 W draw at the battery (at perhaps 2.5 A).
@HKJ tested the Sofirn 21700 cells and arrived at some 17 Wh (down to 2.8 V) at similar discharge rate.
So, according to this logic, and assuming that the heat dissipation sustains it, I predict that this flashlight could maintain 1000 lm for almost 2 hours on the OEM battery. The 2.5 h quoted in the table probably includes some half hour+ ‘reserve’ after the final step-down as Sofirn has done with HS21 for example.