What the paper says is that in that range of concentrations, that the excess of alcohol at the air-liquid interface is enough to extract some of the guiacol from the bulk solution. At higher bulk concentrations of ethanol, there's no particular benefit for the guiacol to separate to the interface (there's enough ethanol in the bulk to keep the guiacol satisfied, so to speak). I forget whether the paper determined if there was a concentration (proof) below which the guiacol did not segregate, but if there was, it would be because the slight increase in alcohol concentration at the interface wasn't attractive enough. In other words the guiacol only is preferentially concentrated at the interface when the difference between the concentrations of alcohol at the interface to the concentration in the bulk makes it worth-while for it. (Worthwhile here means the difference in energy is large enough). Here's an analogy (My family would now roll their eyes, my analogies are lousy). A little kid is deciding whether to get up from his video game and go trick-or-treating. There's no candy in his house, and maybe he'll get a half dozen or a dozen pieces if he hits 20 or 30 houses/doors. Nah, he says, not worth it. But what if he's in a rich neighborhood? If he hits 10 doors, he's sure to get at least 2 dozen, maybe 3 or 4 dozen pieces of candy. Well, if his home hasn't much candy then he'll probably go, but if his home is chock-full of candy (rich neighborhood, remember?) then he may decide it's not worth it either. The choice he makes depends not just on the difference in the absolute amount of candy he will end up with, but also depends on how much he values an extra couple of pieces. It's non-linear, or as Liberals might say "context sensitive". (added: just to be clear, going trick-or-treating is analogous to concentrating at the air-liquid interface, not going is analogous to staying in the bulk phase.)