Duke -
A few things:
First you are correct and Dana is also correct -- if you are beyond the 1 minute or 1:15 point and you don't have a strategy then you should make an educated guess and move on and that is for both Data Suff and Problem Solving.
However, Data Suff is different in that it is designed to not be a choice between 5 answer choices but rather between 2 or 3 choices. This is because you can usually evaluate at least one of the statements in a data suff problem. For example, you are looking for a specific value for x and statement 2 says "x is positive." This is clearly not sufficient, so you can eliminate choices B and D even if you do not understand the question stem or statement 1.
For this reason, if you are behind on time and trying to blindly guess at one or two questions to get back on pace (and yes, the late middle of the test is the place to do this, not at the end) then you would guess at problem solving questions (without reading them - if you are trying to make up time) and not at Data Sufficiency. This is not only because DS can usually be quickly paired down to 2 or 3 choices, but also because DS is supposed to be more like 1:30 whereas PS is more like 2:30.
Your speculation on the placement of experimental questions that do not count toward your score is right on. In order to get good data on their questions, and in order to penalize you for poor time management, the last 7 or so questions are logically less likely to be experimental. Also, you are exactly right about not spending more than three minutes on a question - it may not count!
If I can indulge in quoting myself, here is something that I wrote earlier in the week that I think will specifically help you:
"Many people base their "strategies" on the fear of missing questions. This leads them to the following approach - when they encounter a strange and difficult problem they become focused and alert and put maximum time and attention into this problem. Then when they see a problem that they can get right they try to not only get it right but do so in the shortest possible time so that they can make up the time they just devoted to the really difficult problem.
Here is the flaw in that scenario...between experimental questions that don't count and the chance of randomly guessing the correct answer, a test taker has about a 1 in 3 chance that a blind guess will not count against her. Even if she is able to narrow the problem down to just 2 choices - and assuming that one of those two is the correct choice - her odds are barely better than 50%. In other words, by continuing to devote her time and attention to a problem that continued to stump her, this test taker used three minutes or more and barely improved her chances of getting that question right. (And do keep in mind the very real chance that even if she does get it right that it will be an experimental question and not count in her favor).
Now that she has used too much time on the question that she struggled with she needs to make up that time somewhere - so she will make it up on the next few questions by hurrying so that she can save 20 seconds a question. This is the worst possible thing that she could do - other than just ending the test immediately. GMAT questions are deliberately tricky and are designed to specifically funnel test takers to the wrong choice - even when they fully understand the problem! The test writers want you to try to hurry so that you fall through their trapdoors into incorrect answers - for example, by answering the length of one side rather than the perimeter of the equilateral triangle. Of course the questions are usually set up so that you have to solve for the incorrect choice first and then go through that to the correct choice. If you are trying to "save 20 seconds" then you walk into these traps.
The better way to approach the quant section is this:
When you see a question that you can get right - the devote the attention and the time that it takes to get it right. There is a difference between struggling for 3 minutes to come up with a strategy for a question you find very difficult (this is usually a mistake) and spending three minutes doing what it takes to ensure that you get a particular question (that you know you can solve) correct. It is a matter of approach to the test and it can make a huge difference. Do not try to steal 20 seconds from questions that you can get right only to hand that time over to a question that you will likely miss anyway.
Here are two things to think about:
1) If you do not have a specific strategy for solving a quant problem in about 1 minute and 15 seconds then you will want to choose from the viable answer choices and move on. It is precisely this type of question - where you have no strategy for a long time - that can sap the time and energy right out of your quant section on the test.
2) Set a time standard for yourself - I like to say that after question 25 I should have 25 minutes left (for the last 12 questions). If I am 5 minutes behind, I will just skip (randomly guess) at the next two problem solving questions I see. This will bring me back to the right pace. This is a better strategy than the alternative of " I can catch back up." Be careful here, trying to catch up usually means saving 20 or 30 seconds per question and we already know this is a mistake. Better to sacrifice two or three questions for the good of the test.
Finally, focus on getting questions right rather than on fearing that you might get some wrong. You will get questions wrong. But if you get say, 30 of the 37 right you will be back up to that 44 and this strategy will be one that gives you a consistent score rather than being subject to the swing you saw recently."
Good luck!
