I will write a test in Soapi.CS to see how it goes. C#<->JS is apples/oranges as the browser is pretty good at taking care of those things.
looks like soapi.cs is not handling it well either.
http://api.stackoverflow.com/1.0/search?intitle=C%2523
apparently I am double urlencoding. or maybe i am encoding when I don't need to and the webrequest is encoding it again. I suspect that is the case.
Not sure why you are simply dropping the hash. I will post findings later tonight.