While trying to answer this, I realized there is no answer.
Most of the times I use Spot meteting with AE-L because it makes a lot of sense. I point the camera where I want to meter the mid-tone (actually I do that with highlights and EC+3 most of the time but the idea is the same) and then lock the exposure in order to compose my image the way I want to. This works wonders :)
However there are perfectly good reasons to use a multi-segment metering system (which Nikon calls Matrix) too. This it typical in street photography where you have a scene which you want to meter properly in its entirety (exactly what Matrix metering is good at) but you wish to wait for a subject to enter the scene. If you knew where that subject would be, you could prefocus and then would simply keep the shutter half-pressed. When you do not, you can lock the exposure and only half-press when you can focus on your subject.
Personally, I never used center-weighed metering anymore. I consider it the ancestor of multi-segment because they both look at the whole scene but center-weighed is less sophisticated. So I either use Spot for to read one spot or matrix for the whole frame.
On a Nikon DSLRs what I do is change the center-weighed radius to be the whole frame. This turns Center-Weighed metering into Average metering which is at least much easier to predict. This is good to use when shooting low contrast scenes such as patterns on walls or floors.