It does happen, I think you've got a lot of confirmation bias here, in that we tend to remember negative outcomes.
I've gone from no winter weather advisory to 13 inches once. Most negative busts I've had have been within a range of probabilistic forecasts, i.e. if you looked at the 10% outcome for a low snow output, then there you go. 10% isn't particularly unlikely.
I think it's just the better the models get the more distant we look until we are at a point of relatively low confidence. Modeling has never been better, out expectations never higher.