In this case DF got the same results that AMD advertised, so I don't know where exactly you think they went "wrong". You're just using a single card pairing, using general results, then extrapolating about how a card that's 20% weaker, AND slower, than the 5500xt would do. That to me seems pretty silly.
APIs are a way to interface with the tool. The same code will be understood, that doesn't mean it will run.
Both Sony and MS, working on much more powerful machines, are still working hard to ensure compatibility. A GPU "in the same ballpark" with less RAM and a GPU with less raw power are far more complicated.
Still, why would a 1080p device need to run 4K patches in the first place? It seems like a needless waste of resources for something your target audience won't use.