...then we are imagining a case totally unlike this one, where Google copied the entirety of Oracle's API interface declaration code for the purpose of using it to make money. This is one of the key points Thomas makes in his dissent. So this case is not a case of "we weren't sure whether the material was copyrightable, but in any case our use was clearly for a purpose that would be fair use if it was".
Another point in this respect that Thomas makes is that Google tried four times to get a license from Oracle to use their Java API code, before just copying it and using it anyway. That shows what, in legal terms, would be called mens rea--Google clearly knew the code was copyrightable, because if it weren't, they would not have tried to get a license. So Google now saying they aren't sure the code was copyrightable, as they did in their brief in this case, is, to say the least, disingenuous.