This is all just AFAIK and etc.....

Dialog normalization is an embedded tag that allows the decoder to 'know' where the level of normal dialog is in a soundtrack. This is helpful when trying to maintain a consistent volume level across a spectrum of different audio programs as well as allowing accurate compression of dynamic range if desired, by allowing sounds louder and quieter than dialog to be compressed to a level nearer the dialog level.

If dial-norm is set correctly all movie (and other programs, such as commercials, etc.) dialog should be at the same level while allowing headroom for loud or quiet passages. In practice a lot of engineers don't set the dial-norm value explicitly in the encoder, so it is generally left in a position (the default value) where the decoder will interpret it as a request to reduce overall gain by 4 db.

So, it is my understanding (possibly all wrong) that it is not the presence of the feature but rather the use (or lack thereof) that causes most DD data streams to be decoded at -4 db.