My largely uninformed armchair opinion as to why, is that the author is very performance-driven, and in the end it's very difficult to beat the JVM performance-wise. Lesson: if you want high-perf Clojure, you already have it on the JVM.
Personally, I think there's room for a simple small native Clojure implementation where performance is not top-priority. Small footprint, quick startup, access to native C libs. Still holding out hope for that one.