it is not the density of the basemesh, or the topology, that makes likeness… the angela jolie mesh is very well done, it shows bonestructure and facial muscles already, her planes of the head are well captured, and the mesh has a very good topology on top of that. the other girl mesh, has no bonestructure modeled in, in fact, just the broad outlines are modeled of the face, the rest is hoped here the texture would do. which it does not. never will. the model foundation needs be well, to make a textured head look good.
the arnold mesh is very well done in terms of proportions, that is why a faint likeness is there. the bonestructure however is missing, prominet cheekbones or forms around the mouth,for example, . the topology around mouth will make further work on this mesh difficult.
the likeness of eastwood is not very strong, since there are proportional errors. faint likeness is there because the typical brow squeze he shows is modeled. the model has already quite a lot of polygons but most of them are wasted.
so: the essential is to capture good proportions, and make the mesh as light as possible. likeness comes, as the jolie mesh shows, from good proportions and attention to the correct big forms and planes of the head. not the details… when you have this likeness, you can add details as much as your heart desires, and the likeness will stay… if you begin with a too dense mesh, like the eastwood mesh, it is hard to capture the big forms, and no detail will bring in the true likeness and any changes will be a pain, as too many vertices have to be moved…