Hugger

Turns out, this makes default method is useless for randomisation. I`ve made a datapack that uses 2 structure set, one places red columns with default reduction method, another places blue columns with legacy_type_2. Both sets have the same salt, which should give same result. But default removes lines of structure attempts in Z cardinal direction instead of expected randomisation. It is expected that both columns would spawn at the same position, one over another. Datapack that reproduces it attached below for version 26.2.

The maximum size of the structure can be 257x257 blocks, when max_distance_from_center is set to its max value, which is 128.

The issue you are talking about is not an issue, it works as intended. Fallback pool ignores only size parameter, while you don`t take max_distance_from_center parameter into account, which is always max value (128) for jigsaw ui. Your fallback tries to generate further than that value and thats why it fails.