For update actually does block concurrent SELECT FOR UPDATES, so the article is correct
> Now, other transactions won’t be able to read this row until our transaction is committed or rolled back (for update can only be used inside a transaction).
Finally if transaction isolation is SERIALIZABLE, then a row lock can block normal SELECT reads. This is a consequence of SERIALIZABLE isolation preventing write skew. Transactions can be run in parallel, but there has to be some ordering of transactions that produces the same result as if each transaction was one run one at a time.