Outside of applying the research of persistent immutable data structures (which I think is from the Okasaki paper[0]), I think the most innovative part of Clojure is the fact that it's actually used -- they've managed to build a community and ecosystem with high quality packages and buzz around lisp.
I don't pick Common Lisp/Scheme/Racket despite desperately wanting to because IMO Haskell has a better package ecosystem in my eyes, which is saying a lot. For example, try and find a HTTP server (by far the most common usecase for new projects IMO) that is both multi threaded and async IO capable (i.e. maximally using every core) -- it's surprisingly difficult, I found decent seeming libraries that did either-or, with varying levels of polish and paradigms involved.
I personally prefer CL to Clojure but I just can't imagine trying to convince an enterprise to use Common Lisp these days especially since I can't even really convince myself... Clojure would have the same problem but the ecosystem + buzz is the differentiator in my mind.