This is so naive: at my school they are almost forcing people with 'low' A grades (the inverted commas highlight my annoyance at this situation even arising, where the top grade loses any real distinction) to apply to Oxford over Cambridge. I'm applying to Cambridge (bites teeth really hard in anxiousness) at the moment (may change mind later), and have been encouraged to purely because Oxford have been very ambiguous about how they plan to use LNAT.
Further to the comments raised by JCW, it is undeniable that the vast number of people applying with no chance (and sadly these people generally come from state schools, though not exclusively) dilute the accuracy of the process. The problem Oxbridge finds themself in is when someone applies for the first time from a school; they can't desummon them for fear of turning the teachers at that school anti-oxbridge applications, but often find they simply don't have the necessary grades to apply. What would be better is if they [Oxbridge] actually demanded that candidates declared AS grades. Here however you face the problem of schools which operate under a terminal route, but these tend to be far and few between AND very good schools which do produce good candidates. I think this is definitely one of the objectives of the Supplementary Questionnaire; to weed out those without the AAAA - AAAB prediction.
Regarding skewing the number of interviews, Oxford have made it very clear that their long term objective is to reduce the number of applicants' called for interview to less than 2; thus ensuring that all candidates recieve interviews at two different colleges- as was common practice until the complete abolition of entrance papers. In Medicine, they reduced it to 2.5 apps per place last year, which was the top 320 plus 50 who they would have called to interview the previous year (these were generally people who had completed screwed up on the BMAT). The calculation used is very complex, but accomodates for GCSE performance (1/3) and BMAT performance (2/3) and then factors in school performance (% of A*-A at GCSE, and % of A at A Level) (note that no distinction is made on state or private school; it is purely statistical). This was all done by a central committe. They then checked application forms, to ensure that the personal statement and reference were up to standard (from this 39 people who had the necessary standard were not called up, so another 39 were). They then distibuted candidates amongst colleges. Some colleges, I think LMH was one, only interviewed 4 of their original applicants, whereas over half of Magdalen's applicants were called to interview (but not all at Magdalen). These applicants had two interviews each at two different colleges, and these interviews where scored. Each interviewer was observed on at least five occasions by a member of the Admissions Comittee. Then the scores for all four interviews, the bmat, gcses, personal statement (as an assessment of likely contribution to college life) and reference were placed together, and they gave offers to the top applicants. Again these candidates where then distributed to colleges, if there first choice college was full, but not in a completely randomised way. They tried to ensure a gender and social mix, and if people expressed particular research interests they tried to accomodate these.
The concern raised by people about candidates being 'forgotten' is one I particularly feel- I am considering a college which interviews 120 for Law, each having 2x 20 minute interviews. However, I know (from an Open Day) that they discuss everyone they've interviewed that day in the evening, and rank them cumulatively. So on day one someone could be the best applicant so far, but by day nine, they're applicant twenty seven and not going to get an offer (but will get thrown into the pool!). Conversely, on day someone could be ranked number twelve, but by day nine they're number 16 and still going to get a place. At Oxford they keep people waiting in case they need to reinterview them, but this is laborous and very expensive.