While I have been really unimpressed with how Facebook is handling the fiasco with the Sisters, I'm not surprised. I have spent a long time thinking about the realities of these "real" name policies, and for the most part every reason companies give for having them falls apart with evidence.
Protecting children, either by route of catching pedophiles or stopping bullying, is one of the big ones. I now have a running joke whenever I give a talk about this where I take bets to see how long before someone in the audience asks the pedophile question. Of course, the fact is that if you have a fear for your child's safety on the internet, having them use a name other than their legal name is a great shield against harm. Regarding bullying, while there are many studies that show people who use "online names" use them to bully, I haven't seen any conclusive evidence that using an "online name" leads someone to become a bully. That is, technology magnifies characteristics which already exist.
The whole "quality" argument is so annoying it's difficult to listen to. Aside from the challenge of quantifying quality (and please don't quote Pirsig at me), this is another area wholly without evidence. What it does seem to reflect is a tendency of people to be a bit xenophobic, and scared of things that look strange to them. I think a better way to phrase this is that forcing people to use a kind of name holds them accountable to a certain standard, and sometimes that standard can be used to oppress speech. For many people, using different names has nothing to do with "quality" and everything to do with separating contexts.
The one reason that sticks out as very plausible for these policies is the data broker phenomenon. Simply put, Facebook (and other places) sell your data. Imagine the hypothetical example where Experian contacts Facebook (although more likely, Acxiom), and says they want data. The way the exchange likely works is that they will send Facebook a name, a DOB, and maybe another piece of information. Facebook looks through their database, verifies this, and sends back the corresponding data point if it exists.
Facebook and other places are able to do this because the US laws protecting data are absolute shit. While we do have things like HIPAA in place, places get around that by grabbing data from multiple sources. So for example, let's say your hospital wants to sell your data to make money. They "anonymize" it-- which could be as simple as removing your name from it-- and sell it. What they don't consider is that I can buy this "anonymized" data alongside other data that is less anonymized, such as your Visa purchase history. With these data points, I have one data set showing a hospital visit, and another data set showing a purchase of certain medical devices at the Walgreens next door to the hospital. It really doesn't take much to put these together and figure out what's going on.