...at the high end of the LED's output where it draws a lot of current. At more moderate settings wouldn't the AA's higher mAh give a light more run-time? It seems so judging by the run-times of similar FourSevens 2xAA and 2x123A lights.
The Quark Pro QP2A-X (and boy do I hate their new model name convention) has these specs:
BRIGHTNESS LEVELS Moonlight: 0.3 lumens, 15 days / Low: 2.7 lumens, 3 days / Medium: 24 lumens, 20 hrs / High: 115 lumens, 2.5 hrs / Maximum: 280 lumens, 0.8 hrs
The QP2L-X has these specs:
BRIGHTNESS LEVELS Moonlight: 0.3 lumens, 25 days / Low: 3 lumens, 5 days / Medium: 65 lumens, 11 hrs / High: 160 lumens, 4 hrs / Maximum: 360 lumens, 1.7 hrs
As you can see, the CR123A model has significant output and runtime advantages almost across the board (no output advantage until Medium. I'd expect these were developed with alkalines in mind for the AA versions, so you'll most likely see an improvement there in the real world using Energizer Lithium cells. But it won't make up the difference.