No wait up.
It's appears for me at least trivially easy to be at the network average.

There are two points of variance in p2pool. How many valid shares probability allows your miner to submit in a time period, and how long the block is in a time period.

It doesn't seem to matter too much how far down the MH/s scale you are. If you leave it running 24/7, the probabilities for both work out.

In basic terms. When the mean luck of all the blocks you have processed on p2pool is at 50%, then hopefully! (Because that might take ages!), your mean luck at returning valid shares per time period will be near 50% as well. It's only at that point that you can actually say how much you get on average per day with a degree of certainty.

You can get lucky on shares in a long block even on a slow machine. You can equally miss shares in small blocks. I know. It's a head twister. You need to put faith in the raw probability curves at work and not what your brain tells you might or might not work.

If you have good efficiency and low DOA rates, then you don't seem to get screwed. The stats might be all over the shop, but when you sum up the income over 10 days, you should have a pretty good idea of what you actually get back. Bear in mind, you may pick a period with 4 two day blocks, or 14 short blocks. That's the nature of probability.

The p2pool stats you get from the local webserver are a best attempt at expressing those probabilities. Therefore they are always different to what you actually get for better or for worse. I'm sticking with it now. The end results are so far looking good on the curve to 50% luck.