In all cases (80/20 and 1/99), I would say there is no way to derive an error on the predictions, which make them useless. AT least this is my opinion. Any theory prediction must be accompanied by an error (connected to the assumptions, for instance), and to a way to correctly estimate the error (i.e. the confidence in the predictions). But that's probably my physicist education speaking ;)
And as said above, it is far from being general. All of this is at the end of the day problem dependent.
Yeap. Popper all the way. It needs to have falsification. I am with you. If I saw my article somewhere I would probably rip it apart. My observations are extremely general and take a considerable amount of time to be seen. I just think is better than the 80/20. it makes more sense.
On the picked time scale and the selected example, you are right. The 80/20 does not work everywhere, as the 99/1. At the end of the day, we are in business with empirical stuff that cannot probably be generalized easily.
yeap. I approached it more towards the subjects that it is impossible to gather empirical data.