Read my post above. You can eliminate large sections of the range that have improbable bit patterns. I haven't yet measured how much range that is in proportional to the puzzle length.
How so can you eliminate trillions of trillions of keys? Is this a magic?
How/What method do you use for Range-Eliminating-Measure? That's interesting.
We cannot "definitely" eliminate these ranges, only "almost definitely" eliminate them. I'm relying on statistics mined from private keys of past puzzles and calculate probabilities directly from those statistics.
Eg. There's no guarantee that #66 won't be all ones. But we can be pretty damn sure it won't be. Same concept applies here.