Seeds / Waypoints / Exclusion / Temination masks, all of them must be defined in the same space,
and it is the space where you get the outputs. Lets call it input-output space.
Input-output space is not necessary the same as diffusion space (where you have fibre orientation samples, and brain mask).
But if Input-output space is different to diffusion space, you must provide linear or nonlinear transformations between spaces: