Welcome.
On simplicity
Almost exactly like almost everyone else: “As simple as possible, but not simpler” (Einstein) than their needs.
In my opinion, there are no regular users, only a wide range (e.g. Different ways to structure data) of different users (with simpler users being the majority, but) with different needs (of no meaningful majority). A good thing about Logseq is that it can fit to more than a single level of complexity, thus covering the needs of multiple users (with different levels of quality). And if your needs change, you won’t have to immediately switch to another application.
Only specific points of confusion can be addressed. In short:
If your needs are met with less, don’t be afraid of dropping the rest (Exupéry):
On convenience
What do you mean by “metadata”? Properties? The key word is convenience. What “seems better way” and to whom? Links, tags and everything else than plain text are also metadata. More limited than properties, but arguably more convenient.
No, it is about managing knowledge. But what do you mean by “better”? More powerful queries? If you want to retrieve with queries exactly what you entered, it is better and faster to use directly a database, not a Knowledge Management System (KMS). A KMS promises more insight with less custom queries (thus easier and more conveniently), but doesn’t compete with the raw power of a database. In short:
On the rest
To retrieve, probably not (although it should be possible and should be made trivial). The graph is for discovering things. Keep an eye on How to leverage Logseq’s linked structure?
More or less, they recommend what they have been developing, i.e. variety all around blocks. The recommended data structure is the block. When working with blocks, imagination is the limit. Does this mean that everything works equally well? No, but nobody has explored everything.
A couple (whatever that means) is too little. Keep finding, by both reading and experimenting: