You can also do what I did on the PULL, which is to create a dataset called something like backup, and then replicate into that dataset instead of the root pool. It messes up the expandible directories, but that is just a display issue and since it's a backup I don't really care.