Kyro - what you're saying makes a lot of sense, but you're working under different assumptions than I. Sppose you chose random groupings from the set of all YC funded founders, and give them a random idea from set of all initial ideas, the resulting companies might be different, but the ratio of successes will stay the same. (Assume everyone is equally good friends, etc.)
The launched YC companies all built the company, including all the "atmosphere, communities, ease of use, value, services, etc." in three months or so. Again, any random pairing of YC funded founders could have done this. For example, I don't think any random pair of founders could run a successful MEMS foundry, which requires specific knowledge.
So, now, I read the question as what's different about the founders and not, the company.