Think about it. The VAAC issue one forecast every 6 hours, and if we examine these forecasts since the beginning of the 'ash crisis', let's say 60 days ago, then there is only this one obvious error on May 17th.
60x4=240 forecasts. One of which is wrong, that's 0.4% of the total. Not a bad accuracy level, to be honest.
Forecasts will always be slightly wrong. The forecast you see on something like the VAAC is merely the most probable outcome. The prediction system typically generates a number of possible forecasts and assigns a probability to each that determines how likely it is to occur.
Perhaps on the 17th there were two with very similar probabilities, and the VAAC staff got unlucky and chose the wrong one. Or perhaps they chose one that was very likely, but for some reason one of the more unlikely predictions turned out to be true.