With no autovacuum_vacuum_cost_limit and autovacuum_work_mem set the same as maintenance_work_mem, I wouldn't expect any difference between the performance of manual vs auto vacuum. Still, if you run a manual vacuum analyze verbose, what sort of output do you get? What is maintenance_work_mem set to? Are there indexes on this table that perhaps are very large and needing to be rescanned many times because maintenance_work_mem isn't high enough to handle in a single pass? You might try "create index concurrently, drop index concurrently, & rename index" (reindex concurrently if you were on PG 12) as a sort of online 'vacuum full' on the index(es).
By the way, the best practices for these mailing list suggest partial quoting and responding in-line or below, not "top posting" with the entire conversation below.
