Plot correctly, with an existing program in Python3, a triplot with right values for 2D contours and 1D Likelihood

I have an existing code that produces triplots (that is to say, 2D contours boxes and a diagonal of Likelihood Gaussian functions) : you can see it on the attached figure, which for me is wrong, since the vertical projection of 1 C.L 2D contours on the top corresponding 1D Gaussian match perfectly : this should not be the case. Indeed, when I talk about the vertical projection of 2D contours on 1D Likelihood, I take only the x-component of the semi big axis of ellipses. The problem is set on the following link : [login to view URL] I would like now to have the 1 C.L 2D contours that intersects (when we do a vertical projection) the Likelihood at about 30% of the max Likelihood (which is normalised). I think that in my existing code, the 1 C.L is well computed, so I guess the issue comes from the plotting of 2D contours. If someone could see in the existing code why I don't get this intersection at 30% of max height of Likelihood. I can provide a lot of documentation (like you can also find on the stackoverflow post). I insist on the fact this project is for personal use.