It seems like an unjustified, kind of Panglossian, assumption that all of the unsuccessful research was necessary and that the returns from patents are exactly equal to the necessary costs. Markets are powerful optimization machines, but that is too often taken dogmatically to prove we are in the best of all possible worlds.
I didn't make any such assumption. I just said that if you only cover R&D on success and fail to cover typical failures, people won't have an economic incentive for doing R&D.
In fact the returns are higher than the necessary costs, including the costs of the failures. That is one of the reasons that pharma is profitable.
However what this also means is that big pharma lobbies hard for long patent terms. And when they get them, then other fields, such as software, have to put up with them.
Not unless the R&D cost includes all of the R&D into unsuccessful products that didn't get productized.
Also complicating the debate is pharmaceuticals, where you're also covering the cost of FDA approval, clinical trials, etc. Per https://en.wikipedia.org/wiki/Cost_of_drug_development this cost comes out to several billion per successful drug among established players. (A single drug is several hundred million, but most drugs fail.)