I raised the possibility that the ratty edges of a scan could affect the power spectrum measurements, leading to a scale-free power loss. This doesn't appear to happen... instead, the masking adds some small-scale power and apparently a small amount of large-scale power.
