So, for trans - related issues, especially those situations where a person is considering transitioning, then sex is the only thing that is relevant to GC feminists - because after all, gendered behaviour is not really a thing - after all, what is it to be a man or a woman - either sex is perfectly capable of any characteristics and stereotypically feminine or masculine behaviour? Which is why GC feminists consider there is such a problem with identifying as the opposite sex or transitioning - because what's the point in swapping genitals when there absolutely nothing else that distinguishes a sense of what it is to be a man or a woman?
However, when it comes to things like access to spaces, GC feminists DO believe in gender over and above sex and whatever is between the legs, because they note that there IS more to it than genitals actually; there are characteristics and behaviours that seems to be associated with particular sexes, and it is important to recognise that and maintain separate spaces and treatment at times.
I think there's a bit of a consistency failure between the positions.
Again, you are failing to differentiate between statistical difference and determinitive difference.
There are statistical differences between men and women as a group. The sole determinitive difference is sex. So men are more likely to be violent, tall, brusque, interested in team sports. Some of those are biological differences which will persist regardless of culture, some may be wholly or partly socially produced. For the purposes of this argument it doesn't matter. Regardless of whether it is innate or socially produced it is still not determinitive. It is a mistake to try to apply statistical information about a group to an individual.
Suppose I have a factory that runs a thousand machines. The annual failure rate for the machines is 5%. I can expect approximately 50 failures a year. So I should budget for that. Now suppose I look at one machine individually and I say, "This machine has a 5% probability of failure, therefore it is 19 times more likely to not fail than to fail. Therefore it is unnecessary to plan for replacement or outages" Then I look at each of the other 99 machines and say the same. Now I'm failing at engineering management. Alternatively I could look at a single machine and say, "This machine has a 5% chance of failure". This could be true but only if the chance of failure is uniform. Suppose that in fact the machine has a damaged part. Failure in the next year for this particular machine could be a certainty. That doesn't mean that it isn't part of the group of my one hundred machines for which the probability of failure is 5%. It just means that there is variation across the group.
In the instance of safeguarding and single sex spaces we are excluding all men because as a group they show a higher propensity to sexual violence. That does not imply that to be male is to be sexually violent or that to be sexually violent is to be male. Ideally we would just exclude sexually violent people but there is no practical mechanism to enable that.
Now let's look at your monkeys and just take the study at face value. We see a statistical trend where male monkeys are more likely to be interested in wheeled toys. This does not mean that the 9% of male monkeys who preferred plush toys, or the 18% with no preference are not really male. Both male and female monkeys might prefer either toy or neither. If we had to guess the sex of a monkey based only on it's toy preference we could make use of the information that of the monkeys that preferred wheeled toys 65% were male, whereas for the monkeys that preferred soft toys only 22.5% were, but we cannot infer that a monkey that likes soft toys that monkey is 22.5% male and 77.5% female. It is still either male or female and that is not determined by it's toy preference.