I don't quite understand why having an operation that takes two elements and combines them into a set is particularly useful in describing language. Perhaps it helps in describing how it could be that these strings are generated, but why is this any better than, say, phrase structure rules?
Merge is more useful as a structure building operation than traditional phrase structure rules or X-bar theory, because unlike the latter, it can freely intersperse with movement. If you require structure building to precede all movement processes, you're in some sense postulating at least two different levels of representation. Economy of derivation principles say that one or zero levels is always preferable to two. Furthermore, once we adopt Merge, movement does not need to be a separate, primitive operation, since any instance of movement is equivalent to copying the relevant object and Merging it. This is known as the copy theory of movement. Note that syntactic computation begins once we select items from the lexicon and form a numeration (i.e a set of ordered 2-tuples consisting of the lexical item and the number of times it was chosen (Lexical Item, n)). Since n can be as large as we wish, it follows that the copy theory of movement does not introduce a new operation into the system. Economy principles are observed. There's much more to be said about our Lord and Savior, Merge. We are humbled by His Recursiveness.