5 kts seems to be an awful lot of speed error for an IRS to me.
Is there also position data available from IRS 1? Does this position (and the change thereof) reflect the position of the aircraft and can the shape of the taxi path be recognized? If a large speed bias is indeed present, the shape of the taxi path as represented by the IRS 1 position must be misformed.
What was the IRS 1 speed when the aircraft was static (e.g. parked at the gate)?
Is there other positition data available on the QAR, e.g. from the GPS?
Or if you are really persistent: are there recordings from ADS-B / MLAT / SMCGS.
Groundspeed can be derived quite accurately by fitting a curve through position reports (given that the position reports are quite accurate).