Why this is, is because there are two totally different models cross-mapped to the 16 types.
Temperament was originally measured in terms of a person's "response-time
delay" and "response-time
sustain".
Long delay became what we know as introversion, and short delay became extraversion.
Sustain is often referred to in terms of people vs. task-orientation or "responsiveness". This tells us how much a person holds on to negative emotions, and thus, how serious or "responsive" they tend to be to others. So this model paired Sanguine and Choleric as extroverts, Melancholic and Phlegmatic as introverts; Sanguine and Phlegmatic as people-focused, and Choleric and Melancholic as task-focused. So
here, we had our original temperament matrix.
Immanuel Kant introduced
perception into the mix, but this paired "opposite" temperaments Sanguine and Melancholic as "Beauty-perceivers", which is basically the forerunner to Sensing. The other two, as low in that scale, would thus become "iNtuitive". So now, the matrix was basically "twisted" along the lines of S/N. After this, you would have Kretschmer's system, and eventually Keirsey's, which would be mapped to the 16 types. However, Linda Berens would discover
another set of "four temperaments" in the 16 types (which she called "Interaction Styles"), more closely fitting the original delay/sustain factors of I/E and people/task, which she then named
Informing/Directing.
Keirsey's temperaments would use S/N and another new factor called Cooperative/Pragmatic, which would roughly tie in as another kind of response-delay. But since that matrix was basically twisted by the perceptive factors, the true counterpart to response-sustain would be discovered by Berens, tying together "opposite" Keirsey temperaments (NT/SJ and SP/NF), and dubbed "
Structure/Motive". So you have two four-temperaments systems blended in each of the 16 types. One is more about social skills, and the other is about "action" or roughly, leadership skills.
MBTI used its four dichotomies, and only E/I happened to correspond with one of the temperament factors. Keirsey's "twisted" temperament matrix mapped pretty well (though not symmetrically) to the other three dichotomies. Which meant that the old delay/sustain factors would
not map consistently to any of the other three dichotomies. However, the two "sustain" (people/task) factors would
mirror each other, both using T/F and J/P, but swapping roles according to S and N.
So for a Sensor, T defines "directing" while J defines "structure-oriented"; and F defines "informing", while P defines "motive-oriented". But an iNtuitor switches this, with the T defining Structure, and the J defining Directing; and the F defining Motive, and the P defining Informing. So we see here that T and J, and F and P are basically
interchangeable in defining those two factors, thus, directing and structure are counterparts of each other, as are informing and motive orientation.
So while the SFJ's are informing (S+F) unlike all the other J's, they are also structure-focused (S+J), which does give them a bit of a critical edge, similar to "directing". The STP's are directive (S+T) unlike all the other P's, but instead, they are motive focused (S+P). You can also look at it that any extraverted judgment will be either directive, or structure-focused, or both (in the case of the TJ's). That makes sense, as they are defined as "ordering" the outside world! And that's why on the S side, only the SFJ's and STP's are "out of place" as informatives or directives, respectively. STJ's are just as directive as NJ's, and as Structure focused as NT's by virtue of being both T and J.
Here I explain this and the "two-level matrix" temperament system:
ERICA vs EISeNFelT