Put a slightly different way, the scientific mind is better off learning to walk before it tries to fly. There is little point in asking why twins are linked, because there is no real evidence to suggest that they are. Go out and find that first -- beyond "mere" anecdote, I mean -- and only then, once such evidence is firmly established, is it worth asking "why".
Indeed, it may even be that gathering such evidence helps to answer the question in ways you couldn't have imagined before actually bothering to check if the question was sensible. It's hard to think of meaningful examples, but perhaps you would find other features of twins that appear to play a role in strengthening or weakening the supposed "link", or perhaps you would find the exact opposite, and that no other factors can possibly be an influence.
Regardless, there is as much scientific sense in the question "why are twins linked?" as there is in the question "why is the Universe shaped like a doughnut?" This last is, in fact, not a stupid question per se, because you can easily model a universe with a doughnut shape. But there's no evidence to show that *our* Universe is like that (and, indeed, plenty of evidence suggesting that it isn't), so there's no reason to ask why it would be so.