> all Lisps seem to be more or less interchangeable when it comes to the language
Definitely not true.
It may be true to an extent for implementations of one particular lisp, like Common Lisp or Scheme, but even then there are very real, and significant differences. Simple and regular syntax makes it actually much easier to build many different semantics, and that's what lisps are about.
Racket has less libraries than Clojure and no access to JVM ecosystem. On the other hand it's FFI capabilities to C are very nice and easy to use. Racket macro system - syntax-parse (with syntax-rules and syntax-case and... defmacro for simpler cases) - is the most powerful there is right now. Combined with access to the reader and ability to make reader macros this makes Racket much easier to extend and transform than Clojure (by design). Which takes us to the biggest advantage of Racket, which is it's actually a family of languages, both s-exp based and with traditional syntax, built for specific purposes, which you can mix and match easily. Typed Racket and lazy variant, and FrTime, and Datalog - and more - are examples of this. Then you have module system which is a game changer for unit-testing, object system which is more powerful than most (besides CLOS, but there's always Swindle), the most advanced contracts system among all PLs, and of course quite fast JIT compiler and many more.
Clojure is simpler than Racket, but provides a set of highly opinionated defaults which people are comfortable with. I think Clojure and Racket are like Sublime Text and Emacs - Emacs is strictly more powerful, but it requires much more work to use this power well, while Sublime works well from the get go. But in the end both are infinitely better than Notepad and sooner or later you will come to know both, which is what I suggest you should do.