We've noticed (years ago) the same thing. We went crazy for awhile after installing MOH devices and testing them with our cell phones---we thought something was wrong with the machine, the connectors, whatever.

I did some research back then, and found that someone said that it had to do with the "sampling rate" of the MOH---but the "codecs and processing routines" of the cell phones seem to make sense. Of course, "the bandwidth of the voice path of cellular networks and distorting when the sound clips at peak amplitude" is the best....so far!!