I recommend:
Gary A. Morris, Nathan Harshman, Lee Branum-Martin, Eric Mazur, Taha Mzoughi, and Stephen D. Baker
"An item response curves analysis of the Force Concept Inventory"
Am. J. Phys. 80, 825 (2012) http://dx.doi.org/10.1119/1.4731618
What it says about the FCI questions is interesting, but the
graphical /technique/ is even more interesting.
The idea is that if one of your questions has a score that is
not well correlated with how well the student understands physics
in general, you need to take a close look at that question.
-- It may be that there's something wrong with the details
of the question.
-- It may be that the question is OK but there's something
wrong with the teaching technique for the idea covered by
that question.
-- It may be that the idea itself is just wrong physics, and
no amount of technique can cover up for this.
FCI Q29 and Q15 are particularly alarming examples.
In any case, whenever it is observed that "nobody gets this right"
it behooves us to figure out why.