I'm sorry I didn't get to this until after your test day! You must have posted just after I finished answering questions on Friday and I haven't been on the forums again until today. :(
I'll answer for others who may be in a similar situation in future, though.
The last few days are best used for high-level review of all of your major knowledge and strategies (quant and grammar rules, strategies for the different question types, timing strategies, educated guessing strategies, etc.). Most people don't make a major change in their knowledge or skill level in the final week.
I strongly do NOT recommend doing a practice test a day EVER. That close to the real test, you run the risk of burning yourself out. (Imagine running a practice marathon every day during the week before the real marathon?) The closer you get to the real thing, the LESS you do. (I tell my students no more than 2 hours of high-level review on the day before the real test.)
FYI: CAT exams are really good for (a) figuring out where you're scoring right now, (b) practicing stamina, and (c) analyzing your strengths and weaknesses. The actual act of just taking the exam is NOT so useful for improving. It's what you do with the test results / between tests that helps you to improve. It's pretty much never useful to take a test more than once a week - and that's only when you're at the very end of your study. Before that, 3-4 weeks is appropriate.
Re: "way too tough" questions. Depends on whether I'm behind on time. If I'm already more than 2 minutes behind and I encounter a ridiculous question, I guess immediately and move on. If my timing's okay, then I give myself 1m to try to figure out how to find the right answer. If, after that 1m, I don't really know what I'm doing, I spend up to another min making an educated guess. I DO NOT go over 2m on a question like this EVER. (This process, by the way, is my process for every quant question, assuming I'm not already behind on time. Minute 1 = try to find the right answer. If that isn't working, then minute 2 = try to find wrong answers and cross them off.)
for QA 4 min should be the max time you should spend on any question which doesnt get solved easlily.
No, no, NEVER, no, no. No. :) You can go over 2m IF this is a quant question that you know exactly how to do, but it just happens to be a harder question or has more calculations or for some concrete reason takes you longer to do even though you do know exactly how to do it. You can go up to 2.5m. Maybe 3m once or twice in the entire section. That's it. If you don't really know what you're doing, don't go over 2m.
There will be questions which you can answer in 1.5m or 1.75m, yes, but there are going to be a lot that take 2.25 or 2.5, too. 4 is WAY WAY too much, because this is what's going to happen. The next time you see one that you think is easy, you'll say, Oh, I can answer this one really quickly! And you'll push yourself to answer a 2m question in 1.5 or a 1.5m question in 1. And you'll have to do this two or three times in order to make up for the time lost on that 4m question, right? Guess what you're doing?
You're giving yourself the opportunity to make multiple careless mistakes on easier questions in order to have a shot at answering that one hard 4m question correctly. Getting an easier question wrong hurts your score far more than getting a harder question wrong. (And think about it: if you need 4m to answer the question, how good are your chances that you did actually answer it correctly? The longer you go, the worse the odds.)
So, in short: No. Don't do that. :)