I am not sure how this is related to the transparency and color blending we implemented in our rasterization pipeline. Is it the case we chose our arbitrary blending mode in there (in our case the 'over' operation)? In real world do objects get their own blending mode attribute so pipeline can do different color blending modes based on that?
SO RGB and CMYK are called additive and subtractive respectively because they add up to white and black?
As for the subtractive model, I think the overlapping area means that this color is displayed if combine two material that reflects colors in original circles. Thus the middle black area basically is the result of reflecting no light/absorbing all kinds of colors. Does this make sense?