I don't think that the rank-N system is any more expressive then the rank-2 one. The reason is that by placing a polymorphic value in a datatype we can decrese its rank. In this way we can reduce a program of any rank to just rank-2. So it seems that the issue is one of software engineering---perhaps defining all these extra types is a burden? In my experience, defining the datatypes actually makes the program easier to understand (of course, this is subjective) because I find it difficult to parse all the nested "forall" to the left of arrows, and see where the parens end (I susupect that this is related to Simon's next point).
Of course it's easier to define abbreviations for complex types, which is what "type" is for ... However, if you define new datatypes, you have to change your code, and applying dummy constructors everywhere doesn't make the code more readable ... Cheers, Andres