Why don't you show some more code?

I suspect that the reason for your magic efficiency gain could lie in MySQL query cache. If you want to perform a meaningful comparison, you should turn it off in both cases. If you don't, you get a long time for your first run and a very short time for each run after that.

I might be wrong, but try to run your tests in a reverse order, i.e. first the query with a placeholder and then the one without.