Excellent data!
To be fair, I think it should be surprising that the SFT42R throws farther than SFT40 at all, given the larger LES, which generally lowers the maximum achievable power density. Getting +15% intensity from a larger LES should be pretty big news, and having it running so cool (an indication of very high efficiency) is icing on the cake.
Correct: distance is proportional to square root of candela, which in turn is linearly proportional to brightness given a fixed beam profile. One should not expect a linear percentage drop in a square-root relation.
The square root comes from the fact that distance has the dimension of length, while candela and lux are both per-area. (Note that solid angle is simply area on the surface of a unit-radius sphere).