Could you specify what version of AMS/AMMR you are using? The error looks strange - if it not due to post-processing, then it might be due to muscles somehow losing strength for certain timesteps? Do you use 3-element muscles or simple ones? It could also indicate wrapping problems, but it is hard to say without looking at the model.
It may or may not help. If you can manage to do it - parameter identification should improve the kinematics, which in turn may help with the muscles. But how necessary are the Hill muscles for your work? Could you try running without them and see what happens?
Could we come back to you next week regarding the 3E element muscles? Could you share your trial where it fails? We could then have a look here at what goes wrong and how to fix it.
The 3E muscle indeed takes contraction dynamics into account for defining the strength at a particular time step, but it also makes it more demanding on input parameters and proper calibration. Right now I am not 100% sure what the problem is - it could be a few things.
Muscle forces are subject to muscle recruitment optimization, please see here