Robb is correct that
[COUNT 1] <AND> [RANGE Romans]
finds the 115 lemmas which occur only once in the entire text, limited to the book of Romans. Whereas searching for [COUNT 1] with the Range pop-up menu set to Romans finds the 572 lemmas which occur only once in Romans, but may also occur in the rest of the NT.
Thus the COUNT command counts within the range set in the pop-up menu, not the range set in the Range command.
However, your question was about words occurring in Romans but not in the rest of the text, and your search using HITS is correct. Well done!
Now, as to how to search for unique roots...that is another issue. The roots are an addition to the database, like the glosses, not built into it. I think it is quite amazing that Roy has enabled you to search for roots, use them in a Construct, and set agreement by root. He has not
implemented modifying the HITS command for roots (as you can for lemmas and tags). The HITS command uses the list of lemms, words, tags, or Key numbers from the other window.
I think this is the answer to your need, together with the ability to search for all roots, and to sort the Analysis window by Root. So an argument like:
+*@-[HITS=r GNT] <AND> [RANGE Rom]
would work for you but it does not now work
I have no idea when such a feature might be implemented, sorry!