With 100% efficiency a 4A drive current on an XM-L2 would pull 1.86A if the batteries were at 3.87v under load. With a 90% efficiency a 4A drive current would pull 1.86A if the batteries were a touch over 4.3v under load (not possible in this scenario). I’d say he’s not getting a 4A drive current.
I’m basing that on djozz’s chart where the XM-L2 needs 3.6V at 4A which is 14.4w.
If DBCstm’s cells were just slightly discharged we can assume a best case scenario of about 3.9v (loaded). 3.9v*2=7.8v, *1.86A is 14.508W. If we assume a 90% efficiency (I assume that’s generous but don’t know what the best case for LED2001PHR is) then we get 13.0572W output. Based on djozz’s chart I’d say we’re looking at a 3.5-3.6A drive current. Possibly less depending on just how "not fully charged" the cells actually are.
Knowing the loaded battery voltage and LED voltage would help remove a lot of guesswork here.
He’s using a lower capacity battery, but I assume it behaves like this similar cell:
http://lygte-info.dk/review/batteries2012/Sanyo%2018650%202600mAh%20(Red)%20UK.html