I'm going to be using this colour map :

First, here's a ball with a 1-layered material on it. I've mapped the colour texture to the ref(0) channel and it has NO weightmaps at all.
 

This method works great until you start changing the roughness values of your one BSDF layer.
 

In the above image, because we are only using one layer, the roughness value has affected the entire object, including the decals. Again, if this is the look you want, then you can leave it at that. However for the camera, I wanted to have seperate roughness values for the decals. So I had to add another BSDF layer, and start using some weightmaps.

The weightmap I'm starting with is a black + white version of the colour map. The white is RGB 255,255,255 and the black RGB 0,0,0. When using weightmaps, your layer weight percentage values are over-ridden (notice the T next to the layer name,) and Maxwell takes this info from the colour values on the weightmap (white 255,255,255 being 100%, and black 0,0,0 being 0%.)
 

In order to ensure that the textures are affecting the correct parts of the object, we invert the weightmap on the decals channel, so that the decals are now receiving 100% and the rest is receiving 0%.
  

For now, we will add the original colour map back in the ref(0) channels on both layers. If we change the roughness value of the base layer to 30%, but leave the decals layer as labertian, we get this (shiny ball, dull decals) :

By unloading the colour maps, and changing the ref(0) values ourselves, we can remap the colours of the sphere: