A major strength of Clojure is consistency. If (foo X) works, then (foo Y) should work if X and Y are similar in relevant ways.

The negative offset convenience was considered and rejected when the string library was originally moved into Clojure. In order for a feature like this to be on the table, a proposal would need to include an evaluation of all API fns that take indexes, and a plan to either make all of them handle negative indexes, or have a meaningful rule about which ones do and which ones don't. If someone is interested in doing this, please begin with a motivating problem statement in the Design space.

That said, this feels very low priority.