Wide lenses tend to distort things a bit, it easier to make parallel lines point towards a vanishing point. Using longer lenses will help keeping parallel lines parallel. It is evident that the lens used in the original picture is longer than the one used for the reconstruction in blender.
But changing the lens is not the problem (that is very easy in blender just select the camera and change the lens), the problem is determining what lens to use, so that reconstructed geometry, matches the photo correctly.
It is not a trivial issue, or will take you a lot of trial and error to find the correct values, as the size of objects, and distance to the camera, is only valid for a single point in space, given the perspective of a particular lens.
Because of the complexity of those many different variables, it is preferable to do find the lens value and camera placement in fSpy. That is exactly what that program is designed to do. It uses the very same idea of vanishing points to calculate the lens used, and the camera placement in 3D space.
To help you better, you show a screen capture of the fspy project and how you found the lens size there. Don't be stingy with the information
How can I recreate geometry using a photograph?
How to find the best place to set up the vanishing point axes in fSpy?