How do you think a case of “this explains some of the differences in the population, but not a lot” should look?

And that looks potentially fine for an error bar. For a mean estimate, SE=SD/√N , so depending on what error band they used this looks quite plausible.