This graphic suggests that pre-FDA clinical trials still account for about 30% of the cost of developing new drugs: http://blogs-images.forbes.com/aroy/files/2012/04/fda_05t2.j.... Even if you assume that all of the costs after that are allocable to the regulatory process (which is a liberal assumption--it's unlikely that drug companies would need to do zero human trials in the absence of regulation), developing a new drug is still a billion dollar endeavor.
The reason drugs are such a great example of the applicability of patents is because they're the worst case scenario for the economic issue (free riding) that patents address: high capital costs for a product that is trivially reproduced. That dynamic exists independent of the regulatory process, and exists in other capital-intensive areas too (Tesla has a raft of patents on its technology, and Google has patents on self-driving cars, etc).
The reason patents often seem ridiculous in the software context is that by and large they are not a bad scenario for free-riding. The capital requirements for software inventions tend to be very small.
Looking at capital requirements also exposes a weakness in the article, which uses historical data. Capital requirements for new inventions was much lower in the past than they are today. The era of "lone inventors" is largely dead outside the software realm. We live in the era of "big science" where it costs $1 billion to get anything done. Patents are a lot more relevant in the face of high capital costs than in the face of low capital costs.
TOC at http://www.aeaweb.org/articles.php?doi=10.1257/jep.27.1