This article: http://www.martin.chaplin.btinternet.co.uk/microwave.html suggests that 2.4 GHz was chosen for microwave ovens as a NON-optimal absorption frequency, as this allows some of the energy to penetrate past the outermost layers of food!
At 20 kilometers (about 12.4 miles), a distance this ISP is trying to serve some customers at, that 2 db becomes 40db. Overcoming an additional 40 db loss would require increasing the power by a factor of 10,000 times. They're using amplifiers and directional antennas but the fade margin is nowhere near 40 db. And they do have fading problems. Of course refraction/reflection from temperature/humidity inversion gradients and other propagation effects come into play too. It isn't just simple absorption.
[1] http://www.slideshare.net/1ereposition/long-distance-wifi-tr...
The associated absorption graphs do indicate there is a small peak of 2 dB per kilometer around the 2.4 GHz band due to water. (I'd expect to have a larger drop through a wall.)
From: NASA Reference Publication 1108(02) 1987; "Propagation Effects on Satellite Systems at Frequencies Below 10 GHz; A Handbook for Satellite Systems Design; Second Edition" by Warren L Flock.