There is a huge difference between "large parts of it are useless" and "all of it are useless". And large parts of the non-coding DNA are probably useless, unless you're extremely generous with what counts as "function" when examining this.

Yep!

Evolutionary theory says that it would not be preserved if it wasn't being used. If it was never read or relied on, then random errors would accumulate dramatically faster than in coding DNA. Because in coding DNA mutations generally lead to problems and so they're selected against.

IIRC, much of these non-coding sequences (so-called "junk DNA") are preserved in a way that shows they are being relied upon for something.

Another voice of reason.