I have no idea, I'm a clinician, not a bench worker, but the image first coming to my mind is that mismatch is a random target event, so, you can expect it happening with same frequency all over the DNA chain, that btw, one of my acquaintances proved in his minor thesis for chemistry degree, 70s, with mathematical analysis of forces in bonds, Van der Waals,.., that double helix is the only possible shape for DNA.
If bases differ in the building blocks, this equals different sizes and shapes of molecules, perhaps this structural oddity may be a reason for a certain base pair being more susceptible than others to mismatches.
A lot has been done in simulation of spatial structures from its composition, for example, in defining the shape of pouches of receptors where the active mediators bind, from the chain of AA that build the receptor, this was used to design molecules fitting the pouch, and for knowing the reasons why of the different variants of same receptor being activated differentially by individual molecules of the same drug family.
How far achievements can go in the field of mathematical and geometrical drug analysis and design is something I ignore, but in the time I worked in Pharma Industry, Clinical Research, the medical department received a proposal from a Valencia University of a computer software able to design benzodiacepines.