This made me think of an even more concrete example, but please excuse my use of the volatile RSS subject ;-)
If one were looking for documents by author (specifically, Dublin Core's usage of author, ), one could quickly come to the idea that you could just search the net for any document that simply contained a matching element. But would that get you what you were looking for? No, because not only would you get documents that were by that author, you would also get documents that refer to that author (which was more than you asked for).
RSS, for example, is all about XML documents that refer to documents by other people. One could fine tune searches so that you could say "match , but not within RSS files" or "within RSS files, use the that is attached to refer to the document by that author." All this is possible because RSS makes it clear that RSS files are about other documents, and don't mean that that author wrote the RSS file.
RDF, on the other hand, makes this automatic because the RSS element has an 'rdf:about' attribute. An RDF tool will know that isn't the author of the RSS file, but rather the author of the piece the RSS describes.
(As I mentioned earlier, a user doesn't need to know why or how this magic works, just that it does work if they use the provided element and attribute names.)