Tests made with only rhythm errors give:
a = 96.2% of played errors have been detected.
b = 96.2% of played errors have been detected and well classified.
c = 5.7% of good notes have been recognised as errors.
Detection and classification are quite good. High value of 
can also be explained by the size factor: when too many errors are made on a small piece, it becomes hard to find where good notes are...