Unfortunately that's not necessarily true. A normal distribution is just vastly different than an exponential distribution. Using the central limit theorem allows you to expect the mean to be normally distributed around your samples, but that's about all.
The reality is that taking an average over an exponential distribution is largely meaningless. There are an infinite number of distributions that would create that exact mean. Some of them not at all profitable for certain apps, others very profitable.
The problem with using an average over this distribution is that it doesn't give you anything to work from. You can't figure out who is profitable and who isn't, or how high up the curve you need to be to be profitable. Taking an average over all of the apps in the app store is like taking the average revenue of every website that tries to make money on the internet. It's a meaningless number.
At the end he compares the low numbers in averages to a single anecdotal example of something that had a 100 Million downloads. On the one hand he dismisses the few highly successful appstore apps but then uses another singular example of a highly successful app to recommend a different platform. Overall the methodological approach is deeply flawed.
I could go on, but ultimately, I found this analysis not very insightful. As with any new channel, it takes some time to figure out what works and what doesn't. Who is going to buy and who isn't. In almost all software the distribution is exponential where the top 10 to 20% make over 90% of the revenue. The same is true of the app store.
It all comes back to fundamentals. Find an unmet or under-served need in the market, figure out how big that market is, figure out how much it's going to cost to fill that need, make sure your app fills the need, and start trying to sell it. Then once you've built up a sales platform, then you start scaling it.