Whilst looking at the RPM plots in the AAIB Report, I find it strange that the RPM decreases so abruptly (twice) in the climb. If you extrapolate from 7000 to zero RPM, this would have taken about 6 seconds, whereas the extrapolated 'spool-ups' where in the normal region of 20+ seconds.
Maybe the graphs are not precisely drawn, but the rate of decrease of RPMs seems to indicate some sort of mechanical abnormality.