Really my hot take on "decentralization" meta is that the fediverse is not decentralized for any useful definition of the term.
The protocols affordances in this regard are interesting and would allow for several different definitions of "decentralized," but really fundamentally it is hard to argue that it is "decentralized" in any _useful, practical_ sense.
Back to de facto vs. de jure: I don't think we can compare the de facto implementation of ATProto to the de jure implementation of AP.