For example, two of the three ESFPs I mentioned previously I had originally typed as ENFP (and they scored ENFP on tests and self-reflection). Now, obviously I am not objective, but it took months of thought to realize my error, and more data than just one day's worth of conversation and testing.
Many test questions dividing N and S actually speak more to ability to conceptualize abstract concepts than preference to. Just as questions dividing F and T sometimes speak to ability to think critically rather than preference.
All MBTI tests that I've seen have confounded preference with ability...and for this reason, drawing any statistical conclusions really only speak to correlations between how people ANSWER test questions and intelligence rather than actual type and intelligence.
Your stance is the biggest problem I see with MBTI, to be quite honest. It's a good descriptive tool, especially to quickly explain big parts of personality. But that's all it is. A dimensionality reduction. To get to the interesting stuff about personality, you have to delve a lot deeper.
Your use is like calling a car a sedan and saying you know everything about it. Or saying that an SUV can't be fast because it's bigger than most cars. Well, how the hell do you know? It might have a six liter engine!