I tend to agree with you Mike that the model is broken BUT......
The model works for your offline friends and facebook is trying to replicate that experience. For instance, if you have 5 friends in common offline, chances are very good that your paths have crossed at some time. Even 2 mutual friends could mean something.
Online, the "friend" model works for regular people who have 100-300 friends. And for the most part, the friends will be primarily localized. When you are a blogger and have access to tons of other communities and social graphs though...it tends to get whacky. For instance, if we were "friends" on facebook, your list of people that they would recommend because my social graph would be implemented into the algorithm. Suddenly, my wife may pop up. Or my best friend.
And then, since I am a blogger as well, chances are good that we would be sharing friends already...which would mean that I would be likely to be recommended by Facebook to you even if you or I didn't know each other existed.
Personally, I don't see a relative good fix for this conundrum. When your social graph expands, new opportunities are there.
@LeoDiMilo You've got a point. I guess it works better for everyday normal people. Would just be nice if they were more targeted instead of repeatedly getting some random people from another country or spammers which I've been getting a lot of lately.