My realistic suggestion is this:
Do an experiment where you assign one BFL single to commercial pool of your choice (say Eclipse, because I know you don't hate BFL/Josh yet
), and another to an independant p2pool payout address and have a competition. See which miner generates the most income after 3 months. This will conclusively confirm or deny your hypothesis that p2pool is an appropriate choice for maximizing mining rewards.
the p2pool.info luck pages show exactly what you are talking about (essentially doing the experiment you are talking about). On September 1, the 90 days average showed p2pool earned 9% more than if they were on a large-scale PPLNS pool. Today the trailing average is 7% less than a large scale PPLNS pool over 90 days. Both numbers are including typical PPLNS fees.
And it isn't a hypothesis, it's a mathematically provable (analytically and verifiable via monte carlo simulations) that the expected value of mining p2pool is greater than that of any fee-based large scale pool. This is true even with increasing difficulty. However, at the current hashrates, it is mathematically provable that p2pool will have higher variance.
But Nasty has a point, nastyfans is more than just making money mining, it's doing things that are beneficial to the network. p2pool is one of the ways of maintaining decentralized control over inclusion of transactions in the blockchain.