The behavior is normal, your workflow is flawed. You can have more than one camera in a comp and cut between them. You can shake the text.
I understand that the behaviour is normal, that doens't solve the issue however. I have tried backing up and redoing the workflow but I can't seem to find a way to have of this occur simultaniously. Alternativly would there be a way to stop the camera with the shake from effecting the background. That way I could leave eveything in the main comp and not have to deal with the other issue.
As Rick said, the problem here is your workfow. It sounds like you're trying to combine the results of two cameras, but you can't - the uppermost active camera in a comp is the only camera that works at any given frame. If you precompose a camera and collapse transformations, the camera in the precomp is ignored.
Imagine if you were in the real world, and you had a big set piece of text in the foreground, and a wall as the background. You want the text to shake and the background to stay put. How would you do it? Not by shaking the camera, because the background would shake too.
If you want to shake the text and not the background, animate the position of the text layer shaking, not a camera. If the text already has animation applied that you don't want to break, create a 3D Null layer, parent your text layer to it, and apply the shake position keyframes to the Null.
Simply link the cameras with expressions. Nothing stops you from just inheriting the rotations of the main comp camera and add the wiggle or whatever you are using to this info in the sub-comp. You may require a bit of basic 3D trigonometry/ vector math to keep the camera in th sub-comp in a fixed orbit position to control the text size while the main cam zooms out, but that's poretty much the only bigger issue here and depending on the actual setup may not even be necessary. Otherwise simply do as the others said - wiggle the text, not the camera.