The location of the camera is only relevant in incident metering as it relates to composition. There is no technical reason that the location of the camera should even be a factor in how you take an incident light reading, except as it relates to what parts of the subject are shown in the frame. (E.g. one generally would measure the light falling on a part of the subject that is visible in the frame.) Think of using an incident meter only in terms of where the light is in relation to the subject, not where the camera is in relation to the subject. They are great tools because they take the camera's location and the composition out of the equation. No need to compromise their usefulness by reintroducing the factors that you have deliberately nixed by choosing an incident meter in the first place.
Point it at the light for which you want to correctly expose - the "main light." You have gone to all this trouble to craft light, or choose a location and time of day, that will sculpt the subject the way you want it, why would you then average that light with the dark side, which you have intentionally made dark? You crafted, or chose to use, the light that way because you want the dark side to be dark. If you don't want it to be dark, then change the fill ratio. Don't average the exposures for the light side and the dark side as a matter of course; it does not make sense. For best results, one meters the main light source, unless in very flat, even light, in which case one could point the dome practically anywhere and get the same reading.
Overexposed negs are not a problem to print down, so it is easy to get by without realizing that this method is a problem. But try the "point at the camera every time" method with positive film, and you are screwing yourself in anything but even light.