Yvonne,
I have thought about this a bit. I am not sure, but it seems to me the reason is that for Enumeration attributes in the base entity and for any attributes in an associated entity, Mendix can easily get a finite list of what the values are (in the associated entity, there is an index in place because of the association). For a string attribute on the base entity, Mendix does not have this finite value list at its disposal, and it would likely be a performance issue to regenerate this list everytime a string attribute on the base entity is used as a search field.
That's just my guess, maybe someone at Mendix can give us some more insight.
Mike