The hypothetical that I'm working with is that the insurer can perfectly price insurance by perfectly estimating the _risk_ or _expected loss_ for each individual customer. They can then pool this risk, take a profit, and everyone wins.
This is what I assumed you meant by "perfectly priced insurance". But I've realized you're talking about the case where insurer has perfect knowledge of the future losses each individual will take. I agree in this case insurance will not make sense. Insurance would be free for most individuals and impossibly expensive for those doomed to catastrophe. (The latter group would be screwed, even if they shared perfect knowledge of the future with the insurer.)