I forgot; have you ever explained why you see NF/NT that way (and it wasn't exactly "the other way around" as you said NF could be either Choleric or Melancholic.
I earlier explained why the Supine was a separate temperament: Five Temperaments test!
All of this is all about the two-factor matrix. For original temperament, it was I/E vs people-task. Interaction Styles directly fits this, so there does not seem to be any dispute there which style is which temperament:
IST/INJ (Chart the Course) Melancholic
EST/ENJ (In Charge) Choleric
ISF/INP (Behind The Scenes) Phlegmatic*
ESF/ENP (Get Things Going) Sanguine
"Expressive Behavior": High=E; Low=I
"Responsive Behavior": High="Informative" (F and/or P); Low="Directive" (T and/or J)
the conative area is the same temperament matrix, but in another area of personal interaction: leadership and action, instead of social skills. This matrix is overlayed over the other one, so that each of 16 combinations shares a temperament from both matrices. So E/I is no longer the "expressive" factor in this new matrix.
Keirsey's Cooperative/Pragmatic fits that better, with the Sanguine and Choleric is "pragmatic", (more "expressive" in action), and Phlegmatic and Melancholic as "cooperative'; which is slower to action, trying to work within either the social structure, or maintain harmony. Then, structure vs motive is the other dimension, with Sanguine and Phlegmatic as more "motive" focused (another kind of "people"-focus), and Melancholic and Choleric as Structure-focused (which is more about tasks).
If NF=Choleric, NT=Phlegmatic, then it would be the Phlegmatic that would share "Structure" focus with the Melancholic. Structure would then represent the conative counterpart to "introversion", since Melancholic and Phlegmatic were both "introverted". Choleric and Sanguine would then share Motive-focus, which would become the conative counterpart to their common "extraversion" (expressiveness). NT would fall into the "low Sublime" category along with SP, which would equate to "pragmatism" which would then be the conative coupterpart to "people-focus", shared by Phlegmatic and Sanguine. Meanwhile, the Choleric and Melancholic would share "cooperative" (which would then equate to Kant's "Sublime" recognition, and the conative counterpart to "task-focus"). But none of this makes sense, now. Choleric is definitely not "cooperative", nor "motive-focused"; and this is perhaps the biggest proof that NF can't be Choleric! For the indirect Control temperaments (eC?wC), what Keirsey basically ended up doing is reversing the "high" scores, and substituting making the Choleric high want(motive)/low express(cooperative), and the Phlegmatic high express (pragmatic)/low want (structure).
A similar problem results if NF is Melancholic. Melancholic is by all accounts a very "structure" (task)-focused temperament, but NF is Motive (people)-focused. SJ is cooperative, but Choleric is definitely pragmatic.
Hence the best fit is:
SJ (Guardian/Stabilizer) Melancholic
SP (Artisan/Improviser) Sanguine
NF (Idealist/Catalyst) Phlegmatic*
NT (Rational/Theorist) Choleric in Control
"Expressive Behavior": High="Pragmatic" (T and/or P) Low="Cooperative" (F and/or J)
"Responsive Behavior": High="Focus on Motive"(F or P); Low="Focus on Structure"(T or J)
* In both matrices, Supine is derived by realizing that Phlegmatic is really moderate in both dimensions (hence why it is a tricky fit for either the "exciteable" NF or the logical NT), and that the true low expressive/high responsive is Supine (which provides the NF with its energy).