The second one. As far as I've been told, Mir was conceived because Wayland had some deficiencies, but AFAIK they've been sorted out. So it's really a matter of saving work now.
Is Wayland better than Xorg? Yes, or at least it will be.
Wayland and Mir are distinct projects made to solve the exact same problem in the exact same problem space.
Usually this is a fine idea, but Wayland could really use the support that Canonical has avaliable, especially since it depends so heavily on graphics driver support, which Canonical can help push for using its association with NVIDIA (the only graphics manufacturer with a completely proprietary driver now).
Notice that Xorg doesn't have any real forks anymore. That is because it a much better idea to focus driver support on one library (graphics drivers are hard enough as it is). Unfortunately, Xorg has some inherent problems that Wayland is designed to fix; so until Wayland is complete and stable enough to replace Xorg, we need all our graphics drivers to target both stacks (Wayland and Xorg). That's already difficult enough without Mir demanding its own attention.
