Well, if you look at the stats, the MX2 candidature has actually increased by around 1000 students since 2001, when it was around 2300. Obviously, the criteria for getting into the course have become less stringent.
The bright and gifted mathematicians will always do MX2 and a slackening of the entry criteria is unlikely to be an "incentive" for the already gifted students to enter the course. So it's likely that those additional 1000 students aren't necessarily the bright and gifted mathematicians.
In that scenario, I think what's happening is that approximately the same
number of students are attaining E4s, but
proportionately fewer are. So I don't think it's that implausible that 75% of the candidature could get < ~50/120. Consider 2001 when 45% of the candidates got an E4 (i.e. ~ 1000). Compare that with 2004 in which 27% of the candidature received an E4 and that gives you a figure of ~950. So as you can see, the number of candidates getting an E4 has actually remained fairly constant....looking at it that way, maybe it's not so unlikely that 70% of the candidature received <50/120 in the exam in 2003, because if you "eliminate" those additional 800-1000 students, then the proportion of students getting less than 90/100 actually remains about the same.
That said, I still think that the 2005 cut-off will lie around the high 60s or low 70s.