Given that the probability of obtaining the first 4 digits of a hex "abcd" is 1/65536, the probability of finding 5 repeated prefixes "abcde" within the same range occurs with at least a 3% frequency.
False. It's basically 0% likely to encounter at least 5 "abcde" prefixes in a 65536 range.
What is true: it is 93.94% likely to not encounter any "abcde" prefix at all in a 65536 range.
Hence, 6.06% likely to encounter it at least once.
But, only 0.19% likely to find it more than once.
Where did you end up with the 3%?
Therefore, if we discard
65536 keys around a 5-digit prefix, we have a
97% probability that the prefix is not within the omitted range.
You have a 100% probability that the prefix is in the range, because it's sitting in the middle of it. It was just found, or is the range not "around a 5-digit prefix"?
One would say that you are mixing formulas that relate to birthday paradox (collide any two persons) wih the formulas that relate to finding a specific person, and calling this pruning as valid. It is not, neither from a probabilistic or logical perspective. But good luck with your experiments!