Both Hivemind and SteemSQL interpret the blockchain differently. A really good example of how some Hivemind queries are useful over SteemSQL is doing tag queries. With Hivemind, you can do a query that matches posts that have all tags specified. With SteemSQL, this is difficult.
When I say "all tags specified" I mean that you want posts that match "tag a" and "tag b" as opposed to "tag a" or "tag b".
This is possible in Hivemind because tags are a dedicated table, so doing that kind of query is somewhat trivial. With SteemSQL, you could probably achieve it with sub-selects and temporary tables, but I haven't needed to look into it.
Some more examples: Hivemind Queries