kral,
I think you are misreading something. The first test by far surpassed the other two. In the first test you are using a long string, not a single character.

Additionally, it isn't that ord works better with only one character as an argument - it is that the other two are doing a lot more work. Using substr or split to get the first character is extra operations that are completely un-necessary. The ord function (perldoc -f ord) would work the same if it was a single character or a long string. It is for this reason that the second two tests take more time. They are doing un-necessary work.

I hope this clears things up - L~R