That's what economists generally refer to as the unexplained component. Researchers in the field take the wage gap and they partition it into an explained component and an unexplained component. The explained component looks at gender differences in productivity-related characteristics, evaluated at the male pay structure. The unexplained component looks at the differences in the returns to those characteristics by men and women, evaluated at the female mean characteristics.
Those statistics also include other measures of skill that aren't captured perfectly in a particular study. For example, if we're missing work experience, and we know that work experience influences wages, it then gives us a different estimate of this unexplained component.
The questions related to the gender pay gap are often framed in such a way that it's what the hypothetical woman would earn if she were paid according to the male pay structure. We can use different comparative pay structures, and that will give us a different estimate of this unexplained component.
Depending on the variables used in the study, depending on the methodology used, you could have differences in unexplained components, different estimates. A lot of the studies do suggest that the unexplained component can run between 50% to 75% of the actual gender wage gap, so for lot of our models, if we had better data, the data that we're missing on what determines wages, we could do a better job in explaining the gender pay gap.