1. A quick glance at the graph that Tim links to in his article, shows that it doesn't show error margins or confidence bars. You really can't interpret data without confidence estimates, etc. Tsk, tsk, Tim, should know better! :-)
I wonder if once they are added in the only conclusion might be that Visual Basic is declining (which it certainly is in that graph!). All the other languages are nearly remaining constant, which is interesting in itself.
Java does seem to have the erratic (wavering?) support, though and its interesting how strong C/C++ is despite the brickbrats people hurl at it. And .NET has never really taken off to be the big all-sweeping thing Microsoft probably wanted it to be, it would seem.
2. Book sales, demand by employers and use by choice will all differ.
FWIW, perl is very strong in bioinformatics, although I've come to hate aspects of the language.
There is also a historical element in that many projects are linked to a language for legacy reasons.
Without boring on, I suspect the reality is more complex than book sales alone would indicate.