Hi, Francesco-
By default, 3dReHo will calculate "ReHo" (= Kendall's concordance coefficient) for each voxel in the brain, returning a whole brain map. The ReHo value is over a simple neighborhood of voxels-- you can specify the size of that neighborhood in different ways ("-neigh_* VALUE", "-nneigh VALUE", "-neigh_RAD VALUE", "-box_RAD VALUE", etc.).
If you want to calculate ReHo over voxels that are not in a "simple" neighborhood, then you can upload maps of regions using the "-in_rois MAP_FILE". Within each ROI there (where an ROI is defined as a set of voxels in the MAP_FILE having the same integer value), ReHo gets calculated and dumped out in a text file, the *.vals file, because it isn't a whole brain measure-- it is a per-ROI measure that has been calculated. The radius options and things just refer to the simple neighborhoods above, and have no bearing on the ROI-mapped values-- the neighbood *is* each ROI.
As to why only 237 out of 241 values were being returned, is it possible that the map of ROIs does not fit perfectly on the time series map, and that there are "empty" ROIs? That would be worth checking in the viewer. I am not certain about this point, however.
--pt