Say you have Contact with a nullable many to one on Info, and you write the following HQL query :

SELECT c FROM Contact c WHERE c.name='blabla' OR c.info IS NOT NULL

The generated SQL looks like this :

SELECT c FROM Contact c
LEFT JOIN Info i ON c.info = i.id
WHERE c.name='blabla' OR c.id IS NOT NULL

Which is correct but uses a useless JOIN and confuses potsgresql optimizer.

The following would be better :

SELECT c FROM Contact c
WHERE c.name='blabla' OR c.info IS NOT NULL

Postgresql gets confused and generates a bad plan.