I modified the pitch damping filter for the INPUT mirror to reduce the 9 Hz peak that was showing up in the IR lock accuracy plot.
New filter is ACdamp3 and seems to work better.