If Android users are bi-modal then simply factoring in the number of Android handset sales is not an accurate comparison either.

It sounds like what we really want is a histogram of web usage vs number of users :)