LOL it is not a rule, I was kidding and did not mean to offend. But most will move away from them because you have 20, 30 models being fed purposely alternate (one could argue false) data so close to an event, at what point do you assume the real data will prevail. A lot of forecasting outfits will not refer to them within 3-5 days of an event. When you have such a small spread of 72 or 96 hours, usually the purposely entered alterations do not produce as much effect as say 6-7 days out.