Agreed, but there are a ton of different social graphs out there. If you are looking for data about a person who is only on one or two social sites, the data won't likely be very valuable, but if you are looking at a group of people that all have a many social graphs like on linkedin, twitter, hackernews, quora, that all vary from each other appropriately with the aim of the particular sites, I can see being able to pull out some valuable data. One of the keys is that the users you are interested in actually participate in the different sites differently and have different followers and friends on those sites. If they automatically friend everyone, and crosspost everywhere it makes the social graph useless.