Request to explain in paper

[KSJhon]

I am wondering the meaning the following section in the paper:
"We observe that, since ↑M is low resolution, it can be viewed as containing the low-pass spectral
content of X, which the term↑M−X models.
This frees the network fw to learn a mapping that fuses the high-pass spatial information contained in PAN into X. We input ↑G(M) to the network in order to learn how the spatial information in PAN maps to the different spectral bands in X."
Did you write it wrong?