If one group of YOUR data, we have no idea of the actual content of that data set, has drastically different values of model output then seems reasonable it may be time to delve into your data to see if there is something going on.
Did the proportion of male/female change for that age group? If there is some reason the proportion of gender changes that might affect a model.
Might there be some other confounding reasons affecting that age/gender combination more than others. Perhaps there is something actually related to age to going on. Consider something like "retirement". It might be that one gender in that age group may start retiring before 64 than the other. Or income may change more for one gender in that age group.
Depending on the data I might be tempted to look at different boundaries of age groups.
You may also want to investigate your sampling frame a bit. I know some geographic samples that might have very different results for age because of the proportion of ages is quite different in a location. Look up "Sun City, Arizona" for a moderately extreme case.
... View more