I think a great way to get better with those is to familiarize yourself with types of flawed reasoning used on past tests. There seems to be a finite number of different flaws that they test over and over. For example, if survey results are used to prove something, we should look for reasons to doubt the survey--an unrepresentative or self-selecting sample. Or, if someone makes a causal argument, we might want to look for the possibility of an alternative cause.
Let me expand a bit on lawdog's point here. If you familiarize yourself with the logic that appears most frequently on the LSAT, then you will be better able to spot it when it is used in a question. That translates into you going faster on the test, which is what everyone wants. In questions where you are asked to find an answer that uses the same reasoning, you have a big advantage if you already know the form of logical error that appeared in the argument. All you have to do is then find the one making the same mistake. So if you see an argument making a survey sample error, and you know that it is a survey sample error, it becomes easier to find the answer that has the same kind of survey sample error.
Plus, knowing the typical kinds of logical errors that appear helps you weaken arguments, strengthen them, identify flaws in the argument, and so on. Not everything falls into a nice neat category, but when it does, you can really move fast.
