Unable to replicate figure 2.9 from Krauth: bug in simulation?

[weka511]

[weka511]

Krauth's plot has clear peaks at the boundaries. There seem to be two possibilities:

1. bug in my calculation of impacts;
2. I look at end of run, Krauth samples during run. If ergodic they should be similar, but what if they aren't?