You pick a platform based on market size.
It's not quite so simple, as you point out.
If you target one platform, you target PC's, unless the market for your application is graphic artists, musicians, etc., then you target Macs.
This is the important part. You don't target based on the number of devices sold--the market share--you target based on the platform that your intended audience is using. It is the rare application that will force people to forgo the newest things or switch platforms. There were plenty of DOS developers who eschewed Windows. Where are they now? There were some Mac developers who ignored Mac OS X for as long as they could as well. Eventually, they either updated or disappeared.
Also, as an aside, it can depend on what you're trying to accomplish. Take Bungie, for example, who made a name for themselves on the Mac platform before going to Windows and eventually getting bought up by Microsoft. It was a heck of a lot easier for Bungie to make a name for themselves on the Mac platform than it would have been for them to do it on the Windows platform because there's a lot more competition and it can be tough to shout over the cacophony of other developers. So if I were developing games for Xbox live, for example, I'd be looking at Windows mobile to try to make a name for myself.
I can't buy an android device and get "The Android Experience" - unless you call a balkanized chaos "The Experience".
While I agree, I'd argue that I can get "The Android Experience" from Google's Nexus line of phones. I can get a "Motorola Experience" from Motorola, a "Samsung Experience" from Samsung, etc. This is different from the PC world where everybody has the same "Windows Experience." The problem is that it leads to a commodity environment where all you can really compete on is price in a race to the bottom that nobody wins. Needless to say, Samsung and Motorola don't really want to be in a market like that.