Cambridge research renders speculation about whether "they know what they want/are doing" redundant. For the last five years, at least, they have known that they have been selecting for a characteristic (state school background) that is an independent predictor of poorer performance.
We are conditioned to equate school type with privilege; children in state schools are poor, those in independent schools are wealthy, privileged etc etc. While it is undoubtedly true that children in independent schools are skewed towards the higher socioeconomic end of the spectrum, students from comprehensive schools at Cambridge are even more skewed - 47% of students from comprehensive schools at Cambridge belong in the top SES quintile, compared with 35% of pupils in independent schools.
Furthermore, Analysis of student characteristics and attainment outcomes at the University of Cambridge showed that SES is not a predictor of outcomes, the perception that "poor" students from state schools have to work during holidays, or have problems that do not trouble their "wealthy" peers from private schools, that affect their performance is not supported by the data.