Correspondence between 3D and 2D points using Least-Squares struggles with a simple problem

Question

I am trying to estimate a projection matrix between 3D and 2D points using least-squares method.

I am using a dummy 3D data, from which I am taking the same X and Y and t