The runtimes in this review may be flawed. If you read the forum in light-reviews itself, Mev, the reviewer, stated he thought his sample of the lf3 might have been flawed, because it was not as bright as it should be, illustrating this with the same brightness/runtime graph that he posted in his formal review. There is then a discussion of possibly a flawed reflector, which was not the case and finally, he stated that he changed the driver in his LF3XT and it was much brighter.
However, I don't believe he reran the runtimes, in spite of the fact they didn't look quite as good as the LF5XT and there could have been a problem when he did his runtime tests with a different driver.
In any case, his reviw doesn't state in the summation of badpoints, that runtime was an issue.