@monte Ok, i am sorry to be a bit annoying, but what you qouted what they said, but simply and on your own added the 2.2 part. They refer to a figure in this part.
Statistically, what happens is they estimated a point at which the regression breaks. They do this because they first observe a increasing slope and later no increase at all.
The authors do not say that after 1.62 there is any gain, they very explicitly state several times there is advantage until this point, and afterwards it is flat.
The number you have taken is the upper limit of the confidence intervall they calculated which lies between 1.03 to 2.2. So with the same right one could say there is a gain untill 2.2, one could say that there is no gain at all after 1.03. And the 2.2 is basically also the highest they observe. If you look at the plot, i think it is a streach to see an increase after 1.7 or something.
Sorry for being such a stickler... Ofcourse, you do you and me do me, and i would not tell you to eat less if you want to be super duper sure. But personally 2.2 seems like a lot of optimizing is done in this direction...