[Disclosure: I work for AgileBits, the makers of 1Password]
You should, indeed, be wary of claims of "unsinkability". Quite frankly, anyone who claims that their security is unbreakable shouldn't be in the business. New kinds of attacks get developed and no software is bug free no matter how much we think we've tested and checked.
But there are things that we can legitimately claim. It is often possible to prove mathematically how much work an attacker has to do for a particular kind of attack to succeed. We can prove that many categories of attacks are i
nfeasible where "infeasible" has a technical definition.
So for example, we can mathematically prove that the use of Encrypt-then-MAC authenticated encryption 1Password 4 makes it invulnerable to all chosen ciphertext attacks (CCAs) future and present. Likewise, we can show that the design protects against a variety of other categories of attacks.
What we can't prove is that this protects against yet unknown categories of attacks. Nor can we know that we've implemented everything correctly without bugs. Still, it is useful to know that we can design against unknown attacks of a certain type.
So yes, 1Password is not "unsinkable", but would you rather cross the North Atlantic in a modern designed ocean liner or in a canoe.
You are not alone here. In fact a lot of people who look for or boast about using 256 bit AES keys don't fully understand whether it's appropriate. There are cases were it truly is not appropriate. Take a look at
this (accessible) article specifically about AES128 versus AES256 which will at least answer that question.
Let me quote a little from that article