Screen replacement

Unless there are no other edge details on the device, I never add tracking markers, you just have to get rid of them. The days are long gone where you needed some kind of a tracking marker on a surface to get a decent surface track. Reflections cause problems, lack of detail causes problems, but for the most part, especially tracking a phone in somebody's hand, no tracking markers are required.

The most efficient technique usually involves corner pin tracking or stabilized corner pin tracking. I'll point you to some links. Mocha AE is a great tool for most of this kind of work. I must point out that not all shots are compatible with this approach and the most efficient workflow depends entirely on the shot. Careful planning and execution of the shot can significantly reduce the workload for this kind of a composite. A poorly planned 7-second composite of someone using a smartphone or tablet can take a day to complete, a carefully planned and executed 7-second composite can be done in a half-hour.

You asked about separating the thumb from the screen. Here again, the most efficient technique depends entirely on the shot. There is Rotobrush - often more efficient if you combine motion stabilizing, masking, and pre-composing a copy of the footage. There is manual rotoscoping - again often more efficient if you combine some kind of motion stabilizing or motion tracking. There are Procedural Mattes generated that can be generated using copies of the footage, color controls, channel effects, and other techniques. If the screen has a solid color that is across the color wheel from skin tones, you can even use color keying (Keylight or equivalent). Again, the proper technique depends entirely on the shot.

If you are planning a shoot you need to consider:

• What kind of reflections is going to be coming from the environment?
• Does the light from the screen need to light up something in the scene like the users face or make a glint in the actor's eye?
• What kind of footage are you going to use to replace the screen?
• How are you going to light the scene?
• What should the color balance be?
• Is there going to be a lot of movement in the shot? 
• Do you need to increase the shutter speed in the camera so that you don't have a lot of motion blur to deal with in post-production? (I usually shoot all footage that is going to require separating elements from the background with a shutter speed of 1/125 or 1/200 of a second because you can put motion blur in a shot but it is really difficult to take it out so you can get a good clean mask or matte - even when you use color keying.

If you have an existing shot that you are working with, post a sample or at least one of the most complex parts of the shot, and describe the action in the scene. 

Here are some threads where I explain some of the techniques used in creating composites:

Procedural Mattes

Unfortunately, the two other posts I had saved links for are no longer available. I'll try and find some more posts later.

That might keep you busy for a few minutes. I am creating a training series on compositing that should be ready soon. One of the sections includes screen replacements. I'll post a link when the tutorials are ready to view. I hope this gets you started. 

I personally prefer to do it with black screens and tracking markers. Black screens capture all the reflections of the environmet so the composition is more realistic. Once you clean the tracking markers, use the Screen transfer mode and some color correction (mainly Levels) to compose over your inserted screens. For the tracking task you can use Mocha AE, included in After Effects. It is a must have plugin for this kind of tasks.