So it seems that what you want to measure is co-occurrence - not colocalization. Co-occurrence is simply saying 'yes' or 'no' to the 'presence' or 'absence' of both signals in the same space/area/pixel.
I would think just using the Image Calculater AND operator should do the trick in this case with your individual binary images for each channel. But you would have to test it - would help eliminate some of the 'extra' steps in your current protocol.
Also - on a side... your image is very noisy. In fact, for the latter two channels - can you be sure that what you are seeing is true signal as opposed to noise? Have you checked against some control stainings to be sure?
Hope this helps a bit.