On another note, I don't see why being "soft" has to imply failure to act appropriately in situations that demand such action. It just means that under certain conditions, one prefers to act in ways that are interpreted as more soft and gentle as opposed to hard or cold (or whatever.)
So an individual, let's call him Marcell, can still be attributed the character trait of softness without expecting him to, for instance, back out of a confrontation for fear of hurting the other person's feelings or incurring their negative opinion. It just means that in softness-conducive situations, such as, for instance, a situation in which he is dealing with an emotionally hurt friend who's looking for comfort, he will readily respond in a comforting and nurturing way. Someone who is not soft (at least in the sense I've adopted in my example) would not readily or consistently behave in soft ways in softness-conducive situations.
The point is this:
I think the OP is criticizing our collective tendency to expect traits such as NF "softness" to manifest themselves (more or less) consistently in soft behaviors across situations. An example of this would be the expectation that an NF be a good listener and comfort to his friends in their hours of need, but be ineffectual as an impartial arbiter of justice. The same softness, presumably derived from some enlarged sense of empathy, is consistently manifested in both situations (comforting friends, operating as impartial judge.) Indeed, this kind of thing is littered throughout MBTI literature, with "positive" and "negative" sides of temperament traits listed in order to come up with career advice (e.g., NT's facility with impartial systems makes mathematics an ideal field, but the same traits that make them good at dealing with impartial systems also make them bad at dealing with people, so they should avoid the helping professions.)
I agree that this is more than likely nonsense. There is no reason to think that an NF's occupation as, for example, judge, and the actions demanded of that occupation, would not influence his behavior as much or more than any softness in his personality. Otherwise, how could we ever have competent NF judges?
But I have proposed that one can still have traits attributed to them without the need for cross-situational consistency; there just needs to be consistency in similar situations across time. A soft person is soft because they consistently display soft behaviors in situations that facilitate softness (whew...too many softs in that sentence.) Softness does not need to emerge in situations that don't facilitate it in order to be assigned as a trait to an individual. In fact, the more consistently an individual manifests trait behavior across situations, the more pathological we tend to think of them (behavior clusters are determined to be rooted in pathology the more consistent they are; someone is depressed only if they've shown behaviors that indicate depression in inappropriate situations, which screws up their work lives, for instance, and only if these behaviors have gone on long enough.)
Thus I think that there is a way to consistently say, going back to Marcell, that he is a soft person, but he is also confrontational and aggressive sometimes when he needs to be. And there is a way to view character traits without assuming consistent manifestation in behavior across situations, which helps us to avoid the problems of stereotyping without abandoning the trait assignation entirely.