Interesting... wow.
Reaffirms what I already suspected. the more cells in the series, the more efficient the device is. I knew it wouldn't be great but didn't know how bad it would be. So with 3V, it walks off the cliff in the first 5 minutes and then stays at 10 lumens for essentially forever or 24 hours, whichever comes first.
Just to add something interesting, has there been a chart of M61WLL on 2xAA? If my estimate is correct, 2xAA has the same voltage but greater capacity than 1x123, thus same lumens but even longer runtime. Theoretically.
Another special request: I am scanning the first page, I do not see any M61LL or M61LLL on 2xAA or 3xAA (would be interesting to see both in both configurations. Say 2xAA Alkalines and 3xAA Alkalines and 3xAA NiMH, 2xAA NiMH doesn't make sense I think.
Especially M61LLL (219 or not) on 3xAA.
2 is > 1
3> 2
and of course 4>3