I’m not sure which is the “right” way to do it, but I use 3-point lighting in other 3D software, and I usually set up the positions relative to the object(s) being rendered.
I also have a second camera in each scene, so depending on which camera I happen to be using at any given time, it always seems to work out better if I use the object(s) as the focus for the light setup.
As clarification, I don’t as a general rule do final renders in Blender, as I’m usually rendering what I’m modeling so I can tell how it’s coming along. I then export the model, and import it into the software I’m going to do the final render in. Needless to say, if I was doing a final render in Blender, a very full scene with a lot of objects in it would more than likely need a different setup for the 3-point lights, possibly even needing more than a 3-point light setup.
Just my 2¢ FWIW, as I’m sure others will have other opinions on how they use 3-point lighting.