Since a frame difference would mean different weather/background on every stage of the game. And that would cause each screen to load slower/faster by a few frames to many frames. And that would cause the swinging chains and balls to be in different places.

I have worked on ''demo glitch'' TAS and comparing by framecounter (from the beginning of the second half of stage 5 until the end input on the last boss), it's 477 frames faster on these levels (actually I improved more paths on stage 6, but lost due of a global timer on the chain swings - you need to wait until the rope reaches exactly where is possible to proceed - on the first bathroom room). Another possible improvement for all levels is more lag reduction. Link: http://dehacked.2y.net/microstorage.php/info/62287509/porkypigshauntedholiday-demoglitch.bk2 (almost completely optimized. Some more lag reduction could be possible, I think)

Since this demo % TAS is obviously better than the full TAS, should I remove from consideration the full TAS? Or, is it at all useful archived or something?

I don't know exactly if a full TAS still could be possible to accept, because since the published run is in the vault, probably the ''demo%'' could be considered to a fastest ''any%'' run (I think).