When you say composite, do you mean placing the images side-by-side or actually within each other? The latter probably won't work, as you'll be missing many of the features that the tracker expects in the occluded target.
You can track multiple image targets simultaneously in the Unity extension, by changing the "Max Simultaneous Image Targets" parameter on the ARCamera object. Then you could place targets side-by-side and track off both. You could print them together on a single sheet to keep them fixed, or have them both loose and move them in relation to each other. The child content of each target stays with its parent target as you move them.
Look at the ImageTargets sample project for an example of two image targets in a single scene. You can place them anywhere you want in the Unity scene, the targets will move in relation to each other once the camera views both simultaneously.
Note that using multiple image targets in this way probably won't be more robust for standard size targets. Each target you add adds a bit of overhead to actually locking on a target. However, it might make sense if you were perhaps creating a very large target, say for a tabletop game, and expect that the user will only view a small portion of the gamespace at a time.
One last note: the base SDK does have support for multi-image targets. These are multiple image targets that have a known, fixed position in 3D space. The tracker can optimize a bit more for these, so if you really are fixing the targets in place next to each other this is what you want to use. MultiImage targets are not included in the current Unity extension, but will be added soon!