Sushant
Seeing clearly plays a significant role in your results. look at the guiding assistant run, there are regular excursions of 2" or greater in dec (red line - where no motor is running here) and in some cases much higher in RA:

The aggressive guiding needed for strainwave mounts means you are already chasing seeing to some extent, and the very short exposure cadence exacerbates this. You might try with 1 second exposures to see if the high frequency RMS can be reduced.
I'm somewhat surprised there is any backlash in the Am5N, but apparently it's measured around 700+ms? for a strainwave you might try switching the dec algorithm to lowpass2 with moderate aggression around 60-70. I would also check with other AM5 owners to see what their experience is with backlash and which algorithms they use. afaik it shouldn't have any but maybe this is misleading and we're seeing a function of the resist switch algorithm at work.
I also think your RA aggression is too low, you can see it wandering around a bit. try 0.6 and see if that tightens things up.