Flow forecasts are very useful in determining when we should or shouldn't go fishing based on probable trends, but I find that they tend to be much less than reliable at predicting the rate and extent of rise/drop. It would be interesting to see what data the USGS uses to model their forecasts. I suspect there may be data they don't track that affect these things, so they have something less than the full range of considerations at their disposal. I think ground saturation level at the time of a rain event may be one such attribute. Something else that may affect their forecasts (assuming they do have most of those factors figured in, which is probably the case) would be the duration over which the forecast and observed precipitation fall in a given event. If it all falls quickly, it stands to reason it should drain quickly, producing a steep, relatively complete drop, whereas large amounts falling over longer periods of time should create the slow, steady drops we're accustomed to seeing once the rain arrives in earnest (which I think it's safe to say has happened at this point).
Something else that might affect how quickly flows drop and might complicate forecasting would be changes to the terrain at various points throughout the basin (due to recent logging or a large area within the flood plain being cleared and developed, for instance).
In the end, lest we forget, these are based on weather forecasts, and we all know how reliable those are....