The measure used here is a little suspect. Looking at words used in webpages is a pretty poor measure when many languages have little handwritten language to start with. I just returned from a remote tribal group (The Khasi) in India last month and although their web page penetration may be low, they all facebook or text in their local language. I doubt these social aspects are measured and thus miss widespread use of many languages. Public webpages are often wanted to grow beyond your small group and if a writer knows a more common language, then he would rightly use that to grow the sphere of understanding.