Theres something one can tweak in bitcoin.conf also, try to get Bitcoind GetBlockTemplate Latency to go below 5ms then shares should start to flow in your direction..
5ms would be an amazing feat, never seen it. 20ms is our target.
You also have to take luck/variance into consideration.
How do you get it that low? I've never had p2pool GetBlockTemplate average less that 1 second, and that's on a SSD, but over DSL.