Paul Graham is often saying that the value of Lisp is in large part due to its being so concise. If we have to throw away four lines every time we have a moderately complex arithmetic expression, we're losing quality.
... none of which is to say that Lisp Is Bad (which is certainly not the point I was trying to make in the original article) -- just that the good points are to some degree balanced by some bad ones, and Lisp evangelists don't ultimately help their cause by pretending that even the bad bits are good.