What's to elaborate? You can unfortunately find addresses for which the found prefix (let's say, of length 10) do not all have the same amount of leading bits as the target hash.

In short, this breaks any kind of statistical computations, because some prefixes match less leading bits, and some prefixes match more leading bits. On the outside (base58 view) they will all have the same 10 char prefix. Combined with the 50% chance of having any bit of any hash either a 0 or a 1, this is the explanation for bibilgin's latest mystery quiz: why some prefixes are found more often then others. And it's also just another pitfall in hist theories, once the issue is actually understood.

Thankfully I didn't need to burn tons of kilowatts to find this out the hard way. Every man eats his own yogurt, or whatever...