Can we please please please stop with the 'anecdote of 1 = data' and 'I do it like this and so do my friends (who are self-selected to be as much like me as possible)' arguments. It does not matter what you do, nor does it matter what you think. Businesses have much more data on this than the vast majority of HN readers, and while it's nice to think that they're all stupid in the 'I am so much smarter' sense, it is most likely not true. If Airbnb had, over the course of testing their validation approach against the first let's say 1000 users found out that the majority had little to no overlap in their social graphs between social networks, do you think they would keep using it this way? Clearly the answer is no, which leads to the prima facie conclusion that your generalization of your own behavior to the population at large is flat-out wrong. I'm not saying there is no way it could be that indeed most people have little overlap, just that the evidence we have so far, plus some reasonable thinking, points to the other conclusion.
Of course, if you do have (or can point to) data (i.e., not anecdotes, not 'I think', but hard data, obtained in a methodologically sound way) to the contrary, that would be a valuable addition to this discussion. All these "I don't do it like that so it sucks!" replies are just noise, and that was (in a passive-aggressive way) one of the underlying points in my reply.