That’s the case for correction of chromatic abberation: different wavelengths provide slightly different images on the sensor, but you have no external reference, so you take one channel (green) as reference. Green is used as it is (more or less) in the center of the spectrum, which helps minimising remaining errors.
The coefficients discussed here are for distortion correction. That affects all wavelength, and you have an external reference (straight lines in the image). So you correct all channels.
Of course, you can combine CA and distortion correction in one set of coefficients; in that case, if there is no distortion, there would be no correction for the green channel.