You are absolutely right, and thanks for pointing out that the weights are inverse of the max values. Please bear with me while I explain when this may be useful. I needed to merge/overlay three fluorescent channels (green, orange and red emissions) and wanted to use 'cat' (with false colors). This is something that people often do for imaging different fluorescent reporters in cells. Oftentimes, green fluorescent reporters are much stronger than others, while reds are often quite week. The max/inverse worked in my case because it allowed me to see the week signal by eye, while suppressing the dominant one. So, if someone is merging fluorescent channels, the inverse of 'max' may work to the advantage.
This should not work for images from a color camera. (Of course, I was not thinking about a color camera at all when I saw the original post- I must have a serious case of tunnel vision! Apologies.)