This is a complicated matter, especially when combining different genres…
Level is one thing, and I’d consider the loudest part of each track first. You don’t want a dynamic orchestral track that can and should be played loud, followed by an aggressively maximized metal track that will burn the listener’s ears in ten seconds, obviously. Similarly, quiet pieces (like your “The Last Chapter”) will simply have to be quiet; you can’t master them with the loudest part as reference, because it will sound like the volume is suddenly much louder.
However, that brings is to the next problem: Dynamic range. If you set the gain of these aforementioned imaginary orchestral and metal tracks so that they’re both comfortably listenable at the same volume level, you’re probably going to end up with the loudest transients of the hybrid track being louder than any part of the metal track. (Which is perfectly fine. Some loud orchestral hits are fine, but constant guitar chugging at that same level will be painful.) Even so, this will not work when listening on low volume; the metal track will be fine, but the quiet parts of the orchestral track will be almost inaudible. Basically, it’s all non-linear, and no mix/master will work great on any playback volume.
I don’t think there is a perfect solution for this, unless possibly if you’re in full control of the whole experience (as in, you’re playing it live, through a sound system that you control), but if you really need to mix totally different genres in one reel, the best compromise is probably to nudge them towards a common dynamic range - which means reducing the dynamic range of the more dynamic tracks.
Ideally, one would also strive for a somewhat unified frequency balance and tone of reverbs etc, especially if the tracks are intended to be used in the same context. I suppose for film/game scores and the like, this is less critical than for a music album - or rather, cue mix/master should match the respective scenes and sound design they’re used with, rather than each other. For an OST or demo reel, one might have to do something in between, if the differences are too drastic for “out of context” listening.