This is very telling, and confirms my suspicions both about you personally and about juries and general.
If you think that 90% certainty is beyond reasonable doubt, then that means that you don't have a problem with an error rate of 10%: that is, 10% of all convictions being wrongful.
Are you aware of the sheer number of innocent people you would be locking up with that kind of error rate?
It also means that if I were on a jury and said "I'm going to vote to acquit, because if I voted to convict under these circumstances, I'd be wrong ten times out of a hundred", you would think I was being unreasonable.
I've long held the suspicion that jurors will often convict simply if they think it's more likely than not (>50%) that the defendant is guilty, just because they're not capable of the kind of abstract thinking required to consider acquitting a defendant they believe is likely guilty.
And indeed, it's also very telling that you "still don't quite get it", which suggests to me that you're not used to thinking this way (i.e. in terms of the mathematics of probability theory).
Of course, most of the above is actually somewhat beside the point here, because I don't believe Knox and Sollecito should be acquitted because they're only 90% likely to be guilty; rather, I think they should be acquitted because they're around 99.9% likely to be innocent.