Type descriptions focus NFs as being people oriented, and conventiently ignore the sheer amount of work and problem solving that's required in order to be "good with people". People aren't closed systems in and of themselves. They contian so many variables that'd it'd be next to impossible to know 50-75% of those variables and how they'll affect other variables.
NT's as analysts prefer to work with systems that contain limited variables and can easily be externally controlled.
NF's as "analysts" prefer to work with more complex systems that contain unlimited variables and aren't so easily controlled. (I mean, if it's easy to control, how fucking boring is THAT?)
So NFs have a natural method of dealing with these complex problems and juggling numerous variables, etc.
But that doesn't mean that that natural method is limited to just people. That same method, those same skills, can be used for non-people oriented tasks as well.
By asking the NF to define what they do, you are asking them, in essence, to T it for you. You are asking them to limit it in such a way that for communication purposes, they have to ignore much of what they do and scratch out some of the variables that are involved, etc. Basically...when an NF attempts to describe what they are doing....by default you are going to get a very cheap-ass simplistic "model". But if the NF spends time and energy to create a more accurate model, it's automatically assumed by some people that the NF is actually an NT. Well duh!! If you ask for NT info and they take the time to translate into NT language...then of course it's going to come out sounding like they are NT.