Yes, that is a common “trick”, you convert 3D camera data so that it looks like LIDAR data then you don’t have to change the SLAM software. But the ONLY advantage is not having to change the SLAM software.
There is another use of the vision camera where the most simple explanation goes like this: The robot rolls into a kitchen and vision based software recognizes the front of a stove. Then the SLAM system says “in am in front of the stove”. There is no point cloud in this method.
If you clapse the 3D image to 2D then you are also tosing out most of the information