It would have been good to compare the JS implementation to a C implementation. Given the algorithm, it shouldn't be too bad. I'd suspect that the C version would work much better, but I could be surprised.
Funny anecdote, my first implementation of the algorithm was actually in Go (for concurrency sake). It was an earlier version and wasn't as optimized as the one available here, but I found it actually performed 10% slower at matching nucleotides as compared the same algorithm running in V8. At the time, I attributed it to V8 "writing" better-optimized instructions than I possibly could.
C is one of the few languages that might be more error-prone than JS. I'd be more interested to see a comparison to a safer high-performance language, e.g. Haskell.