I read this study the other day, and it did surprise me. I'm a maths teacher, and I was very aware when I was growing up that I wasn't seen as being as good at maths as my brother was, even though I actually got better results at GCSE, A Level and degree level (he did a straight maths degree, mine was joint honours with computing).
It hasn't been my experience in teaching that girls are seen as worse than boys at maths though. Perhaps I've been lucky in that the maths departments I've been in have been predominantly filled with women (current department has 8 women, 3 men). The departments I've been in have always been scrupulously fair in ensuring that tests/exams are the basis for setting, and we don't quite mark blind but we do mark in such a way that the names are hidden the majority of the time (you can often tell who's paper you're marking by their handwriting though, so it's not a perfect system). Setting is always done based purely on test results (which is one of the recommendations from the study) and if teacher judgement needs to be brought into it for real borderline cases then we usually err on the side of putting the girl into the higher set (as a conscious decision due to us knowing about the gender bias we may subconsciously hold). Nationally, there is still a slight gender gap in maths results at GCSE, which does back up the idea that girls may well be placed in lower sets than they should be. You can't achieve the highest results if you aren't in the highest sets, as you aren't exposed to the same level of maths.
The study also notes that bame pupils are also often put in lower sets than their white peers of equal ability. I can't comment through experience on that one because I've never taught in a school with enough bame pupils to be able to form an opinion. It does seem very plausible though, and is another thing that teachers should be looking out for and trying to challenge their own/colleagues thinking.