this is not possible, coherence requires complex products. But in the screenshot it is shown that he used complex calibration. I would have done the same, but I don’t have an explanation for the strong variations.
About the speed: Such large graphs tend to take much time because of the memory management. You could be faster if you split it into several smaller parts. This was confirmed here Problems obtaining an interferogram from two product sets - #3 by mengdahl and here TopSAR coregistration questions (help!) - #7 by mengdahl