I meant conviction rate from the point of reporting, not from the point at which it goes to a verdict.
That is, still, 6%.
The 6% figure comes from a 2003 (so is not up to date, but I am not aware of any significant changes in legislation since then here) report by Linda Regan and Liz Kelly, both highly respected.
The report compared conviction rates, and practices in European countries, and found that most had archaic systems for rape victims, or worse, kept no statistics as to the amount of convictions (or withheld them).
Countries that came out well included:
Germany
The UK
Poland
The countries you would expect to do badly (Turkey, eastern Europe etc) generally did.
The report's statistics came directly from respective countries justice departments, and the UK department (after a long delay) gave the statistic of 6% from the point of police reporting.
Common issues for stopping the process of a rape trial include:
The victim pulling out so as to put the issue behind them (these trials take ages)
Lack of evidence
Feelings of guilt in the victim
Lack of council
Those that go the whole way (i.e. to a verdict) do generally go the way of guilty, and I don't think anyone is saying that 100% of the time it should be guilty, as, yes, from time to time the allegations are made up, and if proven, the woman who makes the allegations up should do time in prison.
Here is the report if anyone is interested, it isn't too long.
http://www.rcne.com/downloads/RepsPubs/Attritn.pdf