Modeling the head (or hands n feet) separately allow you to concentrate on the form of that one object. You’re not being influenced by the other pieces of the “puzzle”. When you’re trying to create something unique or original I find it helps to start with a blank slate. If I want to do an alien and I start from a human head…it ends up looking like a human head. My brain just sees the human head so my hand sculpts the human head. If I start with a ball of clay my imagination can run with it. I’m looking for form out of nothing, rather than trying to alter a form to fit something else.
You can model the head with Zspheres and get a basic polyflow down from the start if you choose…but chances are you’re going to have to retopo at some point anyway. So you can model from a polysphere/cube and get the form and feel for the character at rather low polycounts and by that time you basically know what the character is going to be and then retopo to get a better polyflow and to build topo where you need it.
Having the tools split allows you to work at much higher poly counts than having the head, hands, feet…etc all in the same tool. That is one great reason to model things separately.
To attach the two meshes you either need to create a new mesh (retopo) and then project the two tools together. You will lose some detail in the process…but at 16million you’re not losing too awful much. You can also mesh insert the two meshes together and then blend them seamlessly. It all depends on your workflow.
Also, having natural seems in a character is very helpful. A watch, a belt, boots, pants, any kind of change in a material will allow you to break the object there and not worry about a seem in the normals (if you’re doing low poly work) because there will naturally be a seem on that area anyway.
Hope this helps.