You mention that you change the "Friction Normal Force Variation". I think this might be the source of the imbalance. By changing those parameters you are changing the tolerance on the trigger to continue iterating for a solution. For example, that particular parameter (normal force variation) is set at 15% by default. That means that if, in the current iteration, the normal load at that friction support changes more than 15% from the previous iteration, the friction force (mu*N) used in the current iteration is off by more than 15% and the program will recalculate that friction force and run through another iteration to get things (more) right. If you increase that variation limit, you might satisfy this check but your numbers drift even farther from the "right" answer.
_________________________
Dave Diehl