If that's right, you may be concluding a priori that natural sciences can't discover such moral differences.

Not a priori. There's lots of history that shows gender roles have been very different in different places and times. Given that background, it seems awfully unlikely that the results of a difficult to test field tend to show that the current gender roles are more compatible with human biology that some other gender roles.

That said, if well-grounded research shows additionally morally relevant differences between men and women, then society should take those differences into account. In fact, morally relevant differences between men and women are readily apparent (e.g. pregnancy). It's just not clear what those differences imply about how society must be structured.