Inconsistency seen when switching to use the reference epoch mcmc files in nominal model

After copying over the mcmc hdf5 files from the first report in each epoch, the unc calculation uses the epoch reference mcmc files. This seems to cause some issues when we try to regenerate the unc in old GPS times. Below I attached 2 plots:

1. the original unc results at GPS time 1381733431
2. using the mcmc files from the first report in the sensing and actuation epochs. The second one is wrong, potentially due to some incorrect kappa applications (unsure).