Why not use a good old standard database in such cases?
Am I missing something?
No; you understand it perfectly well! It's nothing more than a particularly strange form of data integrity assurance (that, as you quickly realize, is trivially bypassed if not distributed across multiple nodes). If you're not using it as a resistance to adversarial peers, and you're not using it for consensus-building purposes, then there's little benefit to be wrung from prefacing each $\text{block}_n$ with $H(\text{block}_{n-1})$.
Even distributed systems such as highly-available and fault-tolerant hypervisors that have to "solve" the Two Generals Problem accomplish their task perfectly fine with non-blockchain-based consensus algorithms.
(However, as the commenters note, this is much more of a systems engineering question than a cryptographic one.)
I will amend this answer to note that Merkle trees per se can be useful in situations where you require, for instance, the public to be able to easily audit or assure themselves that some log has remained append-only, such as in the Certificate Transparency security addition to HTTPS. But (again a nod to Peregrinus) these aren't "blockchains"* since they're not being leveraged for consensus: the scheme itself imposes no restrictions on the whole thing being rewritten; all the auditors can do is sound the alarm if they see a fork in the tree.
(*As evidenced by the fact that a quick search turns up a whole host of schemes proposing blockchain-based augmentations of CT.)
About the one question: By "non-distributed" I mean, that companies seem to use blockchain without the distributed ledger concept. So just one blockchain on their local computer.
– DanielG May 09 '21 at 19:25gitis a blockchain to some people, and not to others. It's not a well-defined term and becomes redundant if you allow centralized consensus. – SAI Peregrinus May 10 '21 at 17:25