Yet, it's hard to find any instance where someone would have used a software patent for its supposed purpose: to replicate the innovation. They're utterly useless.
The other benefit is that it enables companies to do innovation where they would not have been able to keep the idea secret in the first place. E.g. a pharmaceutical drug whose chemical composition would be trivial to reverse engineer. In that case, the patent does not serve the purpose of making the knowledge public in exchange for a monopoly. What it does do, though, is incentivize innovation be rewarding the inventor with the "rights" to that idea.
The real problem with patents is when patents are of trivial things that anyone could have thought of without much effort. That is why I don't support any software patents, because it's so much easier to patent trivial things in the software world than the physical world.
Tort law, for example, predates the U.S., but today we have an economic understanding of what purpose it serves (addressing the negative externalities of risk-creating activities). It's perfectly legitimate to look at those modern economic models to decide how we interpret (and evolve) these existing laws. With regards to patents, we can understand them as a solution to free-rider problems. Regardless of what intuitive understandings motivated them in the first place, the laws should be interpreted with the economic model in mind today.