I agree with you, HopefulOldie - the GMAT scoring algorithm is a lot more nuanced than "600 vs. 700". They heavily research the actual difficulty level of each question so that they can firmly differentiate between a 660 and a 670 question.
Part of that research comes in the form of experimental questions, so you'll see at least a few questions that are NOT designed to test your ability level, but rather to test how someone of your difficulty level responds to that question. For that reason, please DO NOT assume that an easy question means that you're not doing well! Keep this in mind - if I want to prove that a question is a 400-level question, I need to be able to statistically demonstrate that:
1) Those scoring below 400 almost always get it wrong
2) Those scoring above 400 almost always get it right
That means that I have to show that question to several 700-level scorers to ensure that it is, indeed, a question that people scoring well above 400 will overwhelmingly answer correctly - otherwise it's not a valid question. So you will see some experimental questions (you won't know which they are, of course) and that will impact the difficulties that you see.
Now, in terms of practicing the top-shelf questions, you're right that there's a lack of those in circulation, which if you think about it is because:
1) There are more 400-600 level questions necessary for this test than there are 700-800 and 200-300 questions. On a normal, bell-shaped distribution, more examinees fall toward the middle of the range than the outskirts, so they have to write more questions to service that middle ground.
2) 700+ level questions are hard to write responsibly - you need questions that someone scoring a very-excellent 690 will usually miss, but that someone scoring 750 will frequently get right. And it can't be for arbitrary reasons...it needs to show, legitimately, that a 750 examinee is a better problem solver than a 690 examinee. That's really tough to do, and so it doesn't make a ton of sense for the GMAT to retire those questions and put them in general circulation as quickly - if you have that question available, you have to use it!
I'd argue that the best way to study for the 700+ questions is to take the hard questions that you do see and ask yourself: How can they make this harder? How can they change the question slightly to elicit the other answer?
To become an elite test-taker, you really have to be able to think like the test-makers, and learn to spot those devices that they use to make the questions tougher. Yes, there's a lack of really good hard questions out there, but you can create more of them by asking yourself follow-up questions to the harder questions that you do see (what if they flipped the inequality? What if "less than" became "less than or equal to"? What if they asked for x instead of y? What if it said nonnegative and not positive?) so that you're aware of all the subtleties that the truly hard questions on the GMAT will contain.