I think you're right about how the author is making an invalid comparison, but I think his point still stands.
In order to make a good comparison, he would say the same things about Tarver and Qi, but he would have to say something along the lines of "Lisp has been extended in functional Haskell-y directions (Qi) and in object-y Smalltalk directions (CLOS). It would be very difficult to extend Haskell in as many dimensions as easily as it is to extend Lisp"
Not that you necessarily want functional languages to have Smalltalk flavors, but I think it was wrong to compare Tarver with the academics like you say. The plus for Lisp is that it makes implementation of language features easier, you still have to come up with them just like everybody else.