In a perfect world, a prosecutor only brings a case if they have enough evidence to have a reasonable expectation of winning. I honestly don't know if 90% is a problem, but the natural knee-jerk "fair" outcome of 50% would actually mean that half the time, a prosecutor is bringing a case that they shouldn't have. That sounds like an awfully high rate of false charges to me.
90% doesn't sound so wrong to me, and I don't think that's where the problem lies. The problem is the breadth and variety of ways in which the prosecutor can bring a successful case against you.
(This is important because if we're going to push for reform, we want to push for the right things, not just kneejerk that everything is wrong every time we see a statistic.)
1) Defendant is guilty and the police did their job right, bringing strong evidence to the prosecutor. 2) Defendant is innocent, but there's a weight of misinterpreted evidence that his defense fails to overcome. 3) Defense is exceptionally shoddy in the face of an average-to-weak prosecutorial case. 4) Corruption.
To disentangle the effects of gains in (1) & (4) on increases in the conviction rate, you need a metric of police efficiency that's unaffected by corruption.
For instance, a country with DNA or fingerprint forensics will have a higher conviction rate than one without DNA/fingerprints, but as a result of solid police work, not as a result of corruption.
[1]: http://www.forbesrussia.ru/forbes/issue/2008-11/7199-nevinov...