The solver itself does not believe that these values have reached a convergence (although they appear to be part the way there). It doesn't automatically make your results wrong (or right if they all do converge). Behaviors like vortex shedding (Vortex shedding - Wikipedia) will tend to inhibit goals from converging because it is a cyclic steady behavior and make it so convergence calculations never are satisfied. You will sometimes optionally and sometimes be forced to overrule this information based on your judgement as the analyst (you will have to explain why to whomever asks though; so have good reasons for doing so).
Likewise, the goals could converge and they don't predict that the results are good. All convergence is telling you is that the goals don't change over time (within tolerance). Again, you will have to step in an make a human judgement on whether you have accounted for enough of the real life boundaries to call the results a good representation of real life.
Thanks for the reply Ryan.