Why does this Neo4j query skips a match?

If I fill an empty database with this data

CREATE (an:Person { name: "Anders" })
CREATE (ces:Person { name: "Caesar" })
CREATE (boss:Person { name: "Bossman" })
CREATE (geor:Person { name: "George" })
CREATE (dav:Person { name: "David" })
CREATE (an)-[:BLOCKS]->(ces)
CREATE (an)-[:KNOWS]->(boss)
CREATE (an)<-[:KNOWS]-(dav)
CREATE (ces)-[:KNOWS]->(geor)
CREATE (boss)-[:KNOWS]->(geor)
CREATE (boss)-[:BLOCKS]->(dav)
CREATE (dav)-[:BLOCKS]->(dav)

And then run this query

MATCH (x { name: "David" })--(y)-->(z) RETURN x.name, y.name, z.name

I get these results

╒═══════╤═════════╤═════════╕
│x.name │y.name   │z.name   │
╞═══════╪═════════╪═════════╡
│"David"│"David"  │"Anders" │
├───────┼─────────┼─────────┤
│"David"│"Bossman"│"George" │
├───────┼─────────┼─────────┤
│"David"│"Anders" │"Bossman"│
├───────┼─────────┼─────────┤
│"David"│"Anders" │"Caesar" │
└───────┴─────────┴─────────┘

The first row proves there can be duplicate nodes, so why doesn't this row get returned?

╒═══════╤═══════╤═══════╕
│x.name │y.name │z.name │
╞═══════╪═══════╪═══════╡
│"David"│"David"│"David"│
└───────┴───────┴───────┘

There is only one self-loop on the David node, so (David)-->(David)-->(David) would require going over the same self-loop twice. But in Cypher paths returned are not allowed to have repeated relationships (they are allowed to have repeated nodes, as you observed).

even easier to see when visualized. It would, as @finbar says, require to go twice over the same edge.