After talking to someone with a better knoledge of supersonic flow, it appears that supersonic flow within the exhaust tube is plausible. Delicately depending on the shape of the tube it is possible to develop high speed flows without loosing a lot of the energy to shocks.
The piston itself, at ~3000 RPM, cannot be supersonic (unless it has a travel or a few meters), so the picture we get is relatively slow but hot, slightly compressed gas accelerating by expansion down the exhaust tubes up to supersonic speeds, and creating thrust. Thus utilizing both the energy invested in compressing it at the inlet and the residual thermal energy of the burning.
There will be however a shock wave at the exit of the tube where the fast gases meet the subsonic surrounding flow, the effect of which is hard to estimate. Perhaps this is one reason these engines are so noisy.
Thanks to the people contributing to the discussion and to gripen for the NACA report. It's always fun to learn new things.
Bozon