The real question is how well would the students accepted to MIT but who chose to go to less selective schools do (not perfect data, but the best we can get from the real world). I'm personally inclined to believe they do just as well as those who actually attended MIT, but I don't have the data to know.
Interestingly, a similar issue comes into play with YC. A surprisingly high percentage of YC accepted companies are successful (for some definition of success), just like a surprisingly high percentage of ivy league grads are successful.
The real question is how much of that success is due to the fact that companies yc funds are intrinsically better than the rest of the field versus how much impact pg's help has. Maybe pg is just amazingly good at picking winners, and the companies he picks would have succeeded with or without his help.
The only way to get a better idea would be to compare the success rates of people accepted by yc who declined to those actually accepting yc's offer. This wouldn't even be a perfect measure though, since maybe there is some underlying factor (media savvy, proclivity to seek help, etc) that causes a company to both accept yc investment and succeed.
The only real way to tell would be to have pg randomly issue offers to half the companies he normally would, and compare the success of companies he issued offers to against those he would have but didn't issue offers to and a random selection of those he didn't plan to make offers to.
For obvious reasons that would be an impossible experiment to run. In this case though, I'm inclined to believe that pg's help does provide more than a "6.4%" boost to your chances of success.