It's not quite true to say reference counting adds overhead "for each access" -- for example, you can easily have a loop which accesses an object but doesn't modify the reference count. Reference counting adds overhead each time someone "expresses an ownership interest" in an object (wording from http://developer.apple.com/library/ios/#documentation/genera...).
(That's not to say your main point is false. I don't actually know which is faster in general.)
This is true, but has to be thought through very carefully in multithreaded environments where another thread might remove an object from a collection, causing it to be deleted while you're still looking at it. Safe looping then requires locks or cloning the collection (which is suddenly more expensive because it will hit all the contents' ref counts).
In summary, safe and efficient multithreaded code is never easy.
(That's not to say your main point is false. I don't actually know which is faster in general.)