Carolin,
> Maybe -rout does not output the R^2 stats for the error in resting state?
Indeed by default R^2 would not be provided for those regressors of no interest. Conceptually at least two issues seem to be involved: 1) remove the default option -nobout from the 3dDeconvolve command, and 2) bundle those regressors of no interest so that you would obtain the overall R^2. I'm not well-versed as to how to address the issues within the framework of afni_proc.py. Hope someone else will chip in and provide a solution.
Gang