At first I shared your frustration, but I glanced at the paper and it looks to me (as a layman) as though the authors did their homework. It was published in Nature after all.

Here's a high level intuitive argument: Imagine that lots of important parameters about the Earth's climate and ecosystems could be modeled using a bunch of differential equations. Right now, due to humans, our current state is deep in historically unexplored territory. To be safe, it makes sense to try to modify all the parameters we can back to where they were in case weird stuff happens in the unexplored territory.