This talks about accuracy, which is about positioning accuracy, but not repeatability, without which accuracy is mostly worthless. They are also often giving unidirectional accuracy numbers, which doesn't account for backlash in positioning (let alone repeatability).

Think of it as (this is not perfectly right but good enough for this explanation):

unidirectional accuracy - I command a position from 0. Where am I?

bidirectional accuracy - I command a position from 0. Where am I? I command a position in the opposite direction. Where am I?

repeatability - I command a position from 0. Where am I? I command us back to zero. Where am I? Repeat. Compare results.

It is true that, given a single directional commanded position from 0, the accuracy is likely to be 0.1mm or better.

But one missing factor in whether that will be true repeatedly is not just how it is driven, but what is being driven. Is it lead screws? ball screws? gears on rack + pinion? Nothing?

If it's nothing, you will have repeatability issues, because both stepper motors alone, and stepper motors + belts, accumulate error through backlash.

It is possible to get a lot of the backlash out using the right types of belts or gearboxes or what have you.

But even at this price point, you probably want some mechanism (helical rack + pinion, ballscrew, etc) that ensures repeatability, to ensure your accuracy will be worthless.