Decompilation doesn't void white room, we know that since the Sony v Connectix lawsuit from the 90s.
This is a great example, but its not entirely clear-cut. There were appeals back and forth and eventually an out of court settlement was reached where Sony effectively bought them out, discontinued the product and shut it down. So, who really won in the end?
Especially as any precedent was only within US jurisdictions and ROM BIOS's remain copyrightable everywhere else in the world.
It is an interesting case, and I'm glad you brought it up as a counter-argument. The problem however is if you look at the issues the court case was contested around -which was largely about infringement on the Playstation branding- its questionable whether Sony's legal team went about their defense in the wisest manner. For instance they didn't attack the previous precedents like Accolade vs Sega which in turn leaned on Activision vs Atari back in the 80's which were cases very specifically about an external third-party using reverse engineering to gain access to their walled-garden software ecosystems.
Sony were aware of this, which is why the copyrighted the ID string that sat at the heart of PS1's copy protection system, so any attempt to do the same thing would immediately run into a legal buzz-saw.
Which is pretty much why reverse engineering a piece of software is a different legal matter than creating an emulator. When Connectix reverse engineered the PSX Bios they worked around the ID String handshake, and by so doing they jumped the trap that Sony had set for them. The problem is if you reverse engineer the game software itself, its analogous to trying to replicate the string, not the code that recognizes it and behaves accordingly.
I hope that explains the difference, because although it may seem like an arbitrary distinction, the function of the reverse engineered software is very significant in legal terms.