I'm quite surprised at that big of a revenue difference
Although I also suspect it follows a power law, so a few apps make lots of money and most apps make little to no money (on all platforms).
Just comparing aggregates, it's hard to tell what differences are thanks to the platform and what's simply because "the average app" on Android is different from "the average app" on iOS (because of review, barriers to entry, etc.). And there's nothing about costs.
Much as I like Android, I bet iOS tends to be the better deal for paid-app developers right now. You have fewer devices to target and no equivalent of Android's Gingerbread situation. ("The Gingerbread Situation" is also a new punk band I'm forming, BTW.) And Apple customers seem to skew a little spendier, though maybe that's changing.
The big reason is Android has marketshare but some Android users don't buy Android because of Android, they buy it because it was a free phone or because it was the only thing they could get it.
You can't look at phone marketshare when simply determining potential app market.
(Spekaing of Gingerbread, I kind of wonder if Google should show some more charts at http://developer.android.com/about/dashboards/index.html -- show share as a percent of download volume as well as device count. Because I bet those old and very-low-end phones on 2.3 are less actively used than their sheer numbers suggest. That is, I bet devs are safer ignoring Gingerbread than it looks ike.)
MSFT just announced their WP downloads stand at 2 billion (a significant delta from the 0.65 billion estimated here) As for units sold, Nokia alone has sold 20 million Windows Phones or so, and there is another 20% of non-Nokia WP phones on top of that.
Finally, Nokia sales in North America and US, is lagging behind other continents. That's maybe from an app sales perspective Nokia doesnot make a lot of bang!.
US is mainly an AppleLand, and app store app sales get huge boost from having a big portion of iPhone/iPad subscribers here.
How much you'd actually expect to make would depend on which cluster you were targeting.
Median and mode are both highly misleading numbers when you don't know what the shape of your distribution looks like.
This article is just ridiculous.
FunFact: half of the revenue from the app stores go to the top 25 developers.