It must explain both. You aren’t allowed to choose which data to look at and which to ignore. A theory must explain every relevant bit of data better than its rivals. Now you are right that, if you look only at humans and chimps, there’s no way to tell where along the branch between them divergence began (or separate creation from some common template happened). For that it’s necessary to add a third species, say a gorilla. It’s just that you can’t ignore the gorilla. You have to deal with that. And adding the gorilla places that common origin about equally far from both chimps and humans.
Not if you propose that they were created with some differences already in place, if those differences follow the distribution observed in mutations.
If it doesn’t explain all observations as well, and some better, it’s not the best supported hypothesis. Thus we should prefer common descent.