Because if you want to crack a private key by the BTC address, the only option is brute-force (e.g. 160-bit security at most).
But if you have to crack a private key by its public key, Kangaroo, BSGS, or even random sampling (b-day paradox) reduces the search to square-root, so e.g. 160-bit public key is somehow 80-bit secure. But since puzzles 81 to 159 (except multiples of 5) only have the address today, then there is no public-key secure equivalent puzzle to the 81, 82, 83, or 84 bits puzzle, and so on. So, brute-force grows exponentially, but the cost to break them is way higher than the prize. If we had equivalent higher public-key puzzles (165 bits, 170 bits) etc. with public key known, than they weren't actually 160-bits secure, but 82-bits, 85 bits, etc.) - the creator moved those funds way before we had Kangaroo publicly available, so the "160+ puzzles are all actually 160-bits secure" did not make sense at the time.
Puzzle 159 with no pub key is way overkill, it's simply measuring SHA256 cracking performance, not EC security. The highest puzzle that would actually measure EC security would have been #256 (128-bits secure).