I asked the same question some time ago and did not get any good answer with respect to a reliable up-to-date source of statistic data. In the end it comes down to your target market.
People who do not update their os recently are not likely to buy apps in high volumes. So even if these devices exist, the fraction of potential customers amongst their owners is most likely much smaller. That makes it hard to justify the extra effort in providing iOS 5.0+ functionaltiy and same functionality for smaller iOS.
On the other hand it may be suitable for you just to omit some functions for older iOS. That of course depends on the nature of the 5.0+ function that you want to use. Testing the curent os version and then deciding wether to call a function/method or not ist not that much of an effort. It needs to be tested though.
On the other hand there may be an interesting part of the user community that does consume apps but did not yet afford a brand new device. There is a number of devices around which cannot be migrated to iOS 5.x. And I personally would be interested in a) how many of these divices exist and b) how many of these are still in use.
If it is a new app then adressing older devices may not be justifyable. Unless of course you address some very special interest group and now for sure that the amount of oder devices is of a significant value.
Just some thoughts. Sorry that I did not have the statistics handy that you were looking for.