I would say the author just tried to redefined "distributed", and ignored those real distributed systems, e.g. Paxos/Raft-based ones.
A system where there is a centralised leader is still a distributed system so long as at least some of the processing takes place on other nodes.
You're describing consensus algorithms.