Even in single instances of trust there can be advantages to using blockchain for those applications:

• Decentralization can give you better uptime/availability of those documents. If the DMVs website or authentication service goes down, documents can still be authenticated since they and/or their signatures stored in a distributed manner. The internet can go down at your bar but if you have a recent copy of a chain, you can still verify somebody's ID.
• It can make them easier to transfer between parties, and creates a digital "paper trail" which can conform to whatever requirements one might have. For example, you could easily require several parties to sign off any time the document is moved or assigned to a new person.
• You can use those documents and their signatures with smart contracts or other decentralized apps. For example, you could sign up for an account at a bank or a platform like eBay using your NFT'd digital ID and the bank could accept it would needing to manually verify if the id "looks fake" or if your blurry phone picture is going to cut it. They don't have to call up the government and ask them to verify it or pay some third party to match your address against their database of known people, etc.
• Maybe you need better transparency in how many documents are issued and (potentially) to whom. Voting systems, for example, are a use case for this. It could be used for shareholder governance structures, etc.
• Blockchains can enforce rules which centralized entities can't, which is important to consider. An example of how this is useful: imagine the government has a digital ID system and it's run in a centralized fashion, which makes sense, because they are the issuing authority right? Now imagine that centralized system gets hacked and an attacker starts printing and authenticating a bunch of fake ID requests. In the time between when this attack happens and when somebody figures it out, which could be hours to days, banks and other entities could be relying on those fake documents and potentially lose millions. An example of a rule a blockchain can enforce is "this ID issuing authority cannot issue in a single day more than 10% above it's daily average of issuances over a six month period", limiting the scope of an attack. One may say "Well, but blockchain can be hacked too!" which is true, but it's less likely because the software for these networks has thousands of eyes on it whereas there may only be a couple system admins approving changes to your state-run ID database. Open source software is more secure than proprietary for this reason. Additionally, a security flaw needs to effect 51% of the network which isn't likely to happen when you have a diversity of software versions.
• Many smart contracts need ways to protect against sybil attacks (ie one person pretending to be multiple). Quadratic funding being used for charity fundraising is a perfect example. By using credentials issued on chain by centralized authorities, they can verify a person is not multiple people. Quadratic funding is an awesome way to fund public goods.

The problem IS the centralized authority. Can you forever trust a government to not artificially inflate or deflate the value of a currency? The whole point was to have a system with no single authority. No single point of failure.

It is, however, not perfect. The volatility, limited number of transacrions per second , and reliance on an incredible amount of energy expense were the largest of these when the original bitcoin concept was created. Some of these issues have solutions, but its still an evolving technology.