When I first took the test about 5 years ago, I got INFP. Then I took it again on several different websites over the years and it varied from INFP, INFJ, INTJ, and ISFJ. I think I based some of the answers on how I was feeling at the moment when I took them, but INFJ fits me the best.
HumanMetrics always typed me as INTJ when I was in my early 20's (although the T/F was always the weakest dichotomy for me), and cognitive functions tests a year or so ago tended to put me as INFP.
Why? Well, as to the INTJ thing, either I AM an INTJ, I'm not that extreme on F/T so it doesn't really matter anyway, or the questions don't get to the root of cognitive processes and are attributing behavioral trends to type when they shouldn't be. As to the INFP thing, I think one can have strong cognitive functions that they're not 'supposed' to have as strong - which I suspect is my case. Or, again, the questions just aren't framed correctly.
I suspect that it's a combination of both, but probably more to do the the questions. I score as having very high Fi on the function tests, when really those questions could be seeing the Ni/Fe combination. Same for several other cognitive functions. I don't think I'm really all that high on true Fi. I've also thought that it's my bias in taking the test as it's difficult after you know more about typology to avoid gearing your answers a certain way.
I know I'm not INTJ, though I love me some NTJs. The funny thing is that I'd forgotten that I'd ever scored as INTJ until I was looking at old response papers from college and I had written a response in developmental psychology about whether I thought my "type" fit for me (the verdict was: somewhat, but not really). I had been typed as INFJ in high school, also pretty close between F/T. I think the bottom line is just that the HumanMetrics test isn't all that reliable, which kind of goes without saying. I know a few ISFJ/Ps who have mistyped themselves as INFJ/P there.