I'm not sure that statistical lying entirely fits here. The general shape of the visual hasn't been challenged, just the location of the percentage markings. Of course, this doesn't inspire confidence in the visual as a whole, but seems more likely to be a typo or technical issue than an actual lie ( I can't see a benefit to any argument by having impossible data points on both side of the graph ).

I think the technical issue is related to the circles and their sizes.

Feels like the best solution would be to remove the percentages, with more of a vague "more female" or "more male"

But we need the (accurate) scale. Are carpenters 48% female and teachers 52% female? Or is it 5% and 95%? Or 108% and -10%?