There's some incorrect logic in there to point out.

"The math to explain this is simple: most Free users won’t give me anywhere near $3.50 worth of ad impressions."

So, a free user is less valuable than a paid user, fair enough. But not 100% of people who would use the free version will buy the paid version.

As he said in the second paragraph, "it’s hard to say which portion of the increase, if any, is attributable to Free’s absence"

While I'm not saying this is for sure, it is a possibility that, while one user with adverts gives him less money than one user who pays, cutting off all the free users could lose more money in advertising than it generates in sales.