Yes. http://github.com/jtoomim/p2pool master is currently just an old snapshot of http://github.com/p2pool/p2pool master. All of my new code is in 1mb_segwit. It looks like that's what you're running now, otherwise you wouldn't be getting the --bench output.

bitlock1, it looks like your node is sorta working now, but not very well. Your mining efficiency is only 73%. This is probably because of the slow CPU. When you're running p2pool with a slow CPU, it can take several seconds for your machine to download a new share from a peer, deserialize it, build a new share template, and send that share template to miners. In your case, it took 4.2 seconds to do the deserialization step, and up to 1.3 seconds to build a new share template. This means that your miners are hashing on an old share for at least 5.5 seconds out of every 30 seconds on average. That alone will reduce your expected revenue by about 18.3%. (There are also a few processing steps that aren't accounted for in my --bench output, which may be why you're seeing 27% in practice rather than 18.3%.)

Can you try mining on a different node on jtoomimnet to test to see if the problem is with your node or with your miners? You can use ml.toom.im:9332. If your node has publicly open ports, it might also be helpful if you PM me your IP address so I can take a look.

Cryptonomist, your CPU is faster than bitlock1's, and is probably fast enough to run jtoomimnet p2pool with pypy and with an acceptable (but not excellent) DOA+orphan rate, though of course faster would be better.

Edit: based on PMs, it looks like Cryptonomist's setup is working okay. jtoomimnet will often choose higher pseudoshare difficulty, especially on startup, and Cryptonomist is using USB miners with a very low hashrate (~80 GH/s), so he's only finding one stratum share (p2pool pseudoshare) every 15 minutes or so. But it's working.