That is a really lousy tutorial. His method of creating the null and the camera does nothing to line up the camera with the scene and verify the track is accurate. The rest of the workflow is also awkward and a bit convoluted. He presented a technique for lining up the shot with the 3D camera that would only work with a very similar shot and a little luck. Unless you establish a surface on a plane that is in the shot there is no reliable way to know where the tracking markers are or where the camera should be pointed.
If you properly set up the scale expression but didn't pick the same trackers he did when the camera was created it is highly unlikely that you would be able to get the 3D layer to properly line up. It was just dumb luck that made his comp work at all.
The right way to line up a 3D layer with a camera is to hold down the shift key, parent the layer to the Camera, then move the layer in Z about the same distance as the zoom value of the camera. When you use Shift + parent to position a layer at the camera the layer will snap to the optical center of the camera and match the camera's orientation. As long as the layer stays a child of the camera you can move the layer in Z and keep it perfectly aligned with the optical center of the camera. You can duplicate the 3D layer as many times as you want and then adjust the z value to separate them. When they are all where you want them to be, remove the parenting and add the scale expression to all of the layers.
Vet your trainers and make sure they know what they are talking about. Don't rely on any single tutorial to work out a technique. If you are having problems with a tutorial you should post a link to the tutorial when you ask your question. I could have solved this for you a lot sooner if I had seen that tutorial.