RRA problem with amputee model

[pouya Amiri]

Hello,
I am trying to run a CMC simulation with a unilateral transtibial model. To reduce the residual, as suggested by opensim documentation, first I run the RRA, where all the joints are actuated with reserves. I started with the xml setting files for RRA of model 2354 distributed with opensim 4.1. When I run the RRA, the residuals are not bad, but both ankle angles, prosthesis and intact side, have large errors (especially the prosthesi side has an error of ~25 deg), and the foot on both sides seems to slip several cm (I am using the measured ground reaction forces for both sides during a whole gait cycle, starting with the heel strike of the amputated side).
I have done the followings to improve the tracking, but no luck:
1. Increasing the tracking weight for ankles (and reducing the weight of the joints with good tracking)
2. Increasing the optimal force of both ankle actuators

The prosthetic ankle is modelled as a linear spring, whose stiffness I set to zero for RRA; but I assum it shouldn't matter, since the actuators are replaced for RRA.

When I run the CMC with RRA kinematics (as I have them), the tracking is good but the residuals are really large. When I use the CMC directly with inverse kinematic results (not RRA kinematics), the tracking is very good, there is no foot slippage, the residuals are better in order of magnitudes than when I run CMC with RRA kinematics, but they are still larger than what they should be. In addition, there is unncessary co-contraction between ankle plantarflexors and dorsiflexors, all of which are operating close to their optimal fiber length.
Should I bypass the RRA and directly use CMC? if yes, how should I reduce the residuals?
Why CMC and RRA have different tracking performance? My understanding is that they are the same algorithm, and the only difference is that RRA uses ideal actuators for all joints to reduce incosistency between kinematics and GRF, and CMC uses muscles as actuators.

Please advise how should I proceed.
Thanks,
Pouya