Anyway, MBTI can be used to help understand people - which is all you need for it to help you manipulate them. It isn't terribly efficient at doing so since it isn't its purpose... This comes down to an often-repeated comment of mine: ask the questions you want answered. If you want to know how to seduce someone, you gain more from looking at a much narrower set of questions - insecurities, for instance. These don't mirror general personality so well. On the other hand, if you want to know how a relationship would develop, I'd look at attachment styles.
However, you do gain a fair bit by observing along the different axises in MBTI. Keep in mind that behavior stems from all 4 (5/10/12/16/20/etc.) measurements that the system uses. So, if you observe certain behaviors, you can generalize them back to what generates them.
A classic example is with I/E - it is generally thought the root cause of this comes down to the reward system. Introverts tend to be relatively immune to positive thinking (although the biology is counter-intuitive and worth reading if you really want to know), and Es are more addicted to the feedback. Outside of MBTI, you get the opposite with 'Neuroticism'... so there are more dimensions to consider. If you keep within what MBTI can do, then it's useful. So, you can generalise that Is will want less interaction and Es will seek it out - not surprising, but quantifiable. In reality, most people are somewhere in the middle and fluctuate, but the bias helps you understand how to interact (ie: influence) with them.
So, MBTI works just fine as a framework, so long as you understand what you are describing. Interactive effects (namely, functions) are better understood in the validated portions of MBTI, and not the other way around - that is, don't read functions and expect to be able to predict people's behavior from them. Instead, break down the behavior you can see into your 'measurement scheme' (like MBTI's 4 buckets) and generalize what the scheme would be. This is important because even within each MBTI axis (like most systems), there are subcategories that correlate together... which means it is actually pretty rare to have all sub-categories roll together perfectly into the higher categories (never mind into interactive effects, like functions.)