Found bug in CSM which could affect the estimate. A very stupid bug: I only checked for positive roots of the fourth-order polynomial equation. Fix is in the subversion repository.
This could improve the realignment results of PLICP, but I have yet to redo the experiments.