I would say that pain is a cost for other reasons, but that when pain conveys information, the information can be a counterbalancing factor. This makes it a mistake to say that the reason pain is a cost is that it doesn't convey information, equivalent to saying that the reason bombing civilians in Afghanistan is bad is that it doesn't kill Osama bin Laden. The reason bombing civilians is bad is because murder is wrong. Killing Osama bin Laden would be a potential counterbalancing factor that might justify bombing the civilians, but lack of Osama-killing is not the definition of the evil of murder.

Good point; I wasn't careful to distinguish between a "benefit" and an "outweighed cost". However, as I mentioned in another comment, akrasia can blur the distinction. For example, what if pain causes me to take the necessary action against a minor health problem before it becomes a major problem, when otherwise I'd procrastinate? My future self would be very thankful.

Yet you cannot view the pain as some add-on attribute here. The displeasure is part of its usefulness. Simply informing me that "hey, you gotta have this looked at soon" isn't enough; what I need is for my short-term goal ranking to agree with my long-term goal ranking.

Consider someone whose only ambition is to collect every Pokemon in the world.

A contrived scenario. There is no such person, nor will there likely ever be. There might be someone whose only stated, conscious goal is to collect every Pokemon, but their biology prevents them from making that their only actual goal.

So you're right that we don't break the bones of Pokemon collectors, but their friends do try make the collector's long-term goals match up with their short-term goals, and exert social pressure to tame the obsession. As in the above example, pain can be good here.