Having used products that bragged about their performance benchmarks (won't name any names), I respect the Datomic team's position on this. Benchmarks lead to gamed results and you still have to test with your workload.
https://groups.google.com/forum/#!topic/datomic/gr9fscuF6oE
https://groups.google.com/forum/#!topic/datomic/NgVviV9Sw8g
It'd be swell if they put up a list of differences, because using the pull query API + clojure really is quite cool -- it's just good to know what you're getting into.