I would not necessarily say that they help us understand ourselves. Research shows that for most personality test rarely correlate with behavior above 9% in shared variance.
Moreover such test as the unvalidated Meyers-Briggs, almost always do not correspond to meaningful components. Research generally shows that rather than it mapping to pseudo Jungian parameters, like most personality tests, it maps more closely to the Big Five personality dimensions. https://en.wikipedia.org/wiki/Big_Five_personality_traits