As time progresses, it's harder to come up with ideas that can truly be called original because of the increasing mass of games behind you. As a result, it's a much safer bet to focus on trying to take something that was both innovative and good in the past and turning it into something even better.
Besides all games are a combination of elements, it's just the particular combination that makes the game unique. You referenced Halo as a rehash in your first post, but that's a load of crap. Sure, you can probably point to a prior instance for just about everything in that game, but it's the combination of game elements that makes it unique. It's the pacing, the setting, the little elements being put together - like recharging shields, vehicles, the particular weapons, and the online - that make it different. It's like designing a building. The materials and styles may have all been used before, but that doesn't mean you can't make a unique creation from existing ideas.
Besides, there's a reason people look forward to sequels. It's the hope that someone takes a treasured game (or one with perceived potential) and makes it better. That's the goal, really. Of course, sequels can be profoundly innovative - look at the progression of Ultima games - but even if we ignore that, the new games at least are upgrades and reconfigurations of titles people love.
Think of Counterstrike Source. It is, in many ways, a carbon copy of the original Counterstrike. People still play it, though, because the original Counterstrike was a brilliant multiplayer game. Getting more of the same but with some visual and minor physics upgrades is all a lot of those players really wanted.
Anyhow, there are so many components in modern video games that minor tweaks that would have been hugely noticable in the 1980s are completely ignored. I mean, remember the differences between Pac Man and Ms. Pac Man? Like how ghosts in the latter had sort of randomized movement? Or how there were new maze styles? That counted as a big deal back then. Now, something like new levels and better AI would be almost completely ignored in the dozens of other facets in a game. Halo and Halo 3, for instance, probably use significantly different AI. The levels are obviously much more varied than in Pac Man and Ms. Pac Man. But that's not the kind of thing we count as different anymore - at least not as a notable thing.
Plenty of good games still come out, at any rate. Probably more than ever before, which really means nothing stands out as much. Really, I don't think you should discount Fallout 3 or Europa Universalis III just because they're sequels. Nor should Halo be considered generic because its gameplay elements are traceable to other sources. On top of that, there are games that use some awesome gameplay elements, but the games themselves may fall a bit flat on the overall. Assassin's Creed is a good example of this, where the character motion and fluidity is fantastic. Something could be certainly done there to incorporate such things into an even better game, whether it be a sequel or just a new game entirely.
This has already become quite a wall, but my point is that games don't have to add an nth dimension to be great. Often times just shuffling up the Rubik's Cube is enough to give you a whole new challenge.