Clearly there's no or very little correlation between similarity of attractiveness levels and relationship stability because ignoring the whole internet dating topic, people have lived in communities ranging in size over at least 5 orders of magnitude if not more, yet actual divorce stats of small rural villages vs megacities are not separated by anything near 5 orders of magnitude. Or whatever other measure of stability you'd like to use, none of them approach the incredible variation in size of the dating pool and therefore similarity in first impression rating.

A physics analogy would be that no matter how strong the theoretical argument, its widely observed that over many human accessible orders of magnitude in mass of the object, gravity makes things fall at the same rate. Leading to the (actually inaccurate) theory that the force of gravity has no correlation with the masses of the objects.

Given that there's observationally no correlation between two variables and not much theory either, then observed variation in one variable is by the above, unsurprisingly not correlated to the other variable.