I agree with you Jim, but not entirely. Innovation makes a game better by making it something new. There are a lot of very flawed games I enjoy just because they are so DIFFERENT from anything else around. Games like mirror's edge or fez or minecraft. I will say though, that while I think that innovation (usually) improves a game for me, it does not necessarily make it good. It is just a single positive point in the game's favor, and can be outweighed by enough negative points.
Minecraft has one of the best core mechanics of any game I've ever seen. However, it is also unfocused, buggy, has a lot of poorly implemented ancillary features, and takes FAR too long to update. Fez has a brilliantly creative core mechanic, amazing art, and a very relaxing, unusual tone. It also lacks direction, has a shitty map system, and its puzzles are VERY obtuse at times. Mirror's Edge has incredibly fun platforming, and some of the best visual design in game history, but its flaws are so numerous that it is not a very good game overall.
If a game is going to lack innovation, it has to do everything VERY well. It has to be polished. I feel that if you are doing something in your game that has been done in dozens of games before you, you MUST do it well or it is not really worth my time. Also, you should be able to learn from these past titles. See their mistakes, what works and doesn't work. Combine these elements to make them perfect. That is when using old ideas is ideal. Blizzard is known for doing this. Starcraft 2 doesn't really have ANY mechanics that haven't been seen in some other RTS before, but it does these things better than almost any other RTS (at least in my opinion.)
The other time that you don't need to innovate is when the mechanic or system is not central to the game. The example you used was JRPG fight systems. They do not need innovation because they are not normally the point of those games. They are simply a tool used to bring across the core of the game - the story. In that case , you do not want people to even notice those systems. They should just be functional. IF you try to innovate, you will just draw attention away from the core of your game.
What would help publishers learn how to innovate properly is if they were willing to create smaller, cheaper games. Fund a game on the scale of Fez or Braid or even Amnesia. That would allow publishers to test out new and unproven ideas without having to worry too much. I sympathize with them. You can't take too many risks when your game costs dozens of millions of dollars to make. You just can't. Those AAA titles should be the super-polished ones, the ones that take proven ideas and perfect them. Smaller titles have room to innovate.