This isn't the right conclusion to draw from his investigating. All he showed was that, as you said, few people upgraded from the free app to the paid one. This does not necessarily mean that the free app did not significantly help the paid app sell. Maybe it was fairly common for one geek to buy the app, show it to his friends, who would then try it because there was a free version and not ultimately ever buy the paid version, but still talk about the app and introduce it to people who did buy the app. There are uncountably many other situations like this that could have happened and caused the presence of a free app to have helped the paid app; this is very far from a controlled experiment.