"XRef Objects” VS Standart Scene, is it an evident victory of the first or just a different method?"

Tried to use the scheme with XRef Objects models while building complex scenes. This led to reduction of number of verticles in the scene more than in 15 times! Autoback has reduced to a few seconds. You may save at least every minute!
The size of the scene, which used to be up to hundreds of MB, reduced to 1-2MB. It's fantastic! Why didn’t I use it before? Why don’t my colleagues use linking, but prefer working with models inside the scene? Perhaps there are some reasons because of which this method does not work in every situation?