I "tested" as an INFP on this one. I don't remember all my percentages and I've already closed that tab on my browser, but it had the I at 22, P in the thirties or so, and the N and F were "distinctively expressed." I really don't know about this test's overall reliability. It's pretty close though, and in fact most of the online tests I've taken result in an INFP score, but I think that's because I am probably not super-intense on my Extraversion, though I do have that preference overall. The descriptions of ENFP fit me better too, though of course they don't fit anyone exactly. Of all the tests I've tried, the Keirsey Temperament Sorter seems the most consistently accurate, and I am a definite ENFP on that.
I think that some questions leave room for error, like the "do you have a neat or messy desk" J/P questions and some of the I/E questions too. I say this because in order to get an accurate typing, a person should answer in accordance with their honest, innate inclinations and not with their current habits or with things that are heavily situational and dependent on lots of factors other than straightforward MBTI functions. Certain questions are hard to answer that way, though, depending on how they are phrased. For example, I would naturally tend toward a disorganized desk, especially when I was younger, but I have made a concerted effort as I have matured to have a neat work environment. So I would answer "yes, I have a neat desk", because the question did not ask what I would do when I was younger or if I did not have work obligations that require me to be neat, or if I hadn't made a conscious decision to be tidy. It asked me what my desk looks like now. So that would be a point for "J" when in fact I have a "P" preference overall. Though the more I think about it, having a P preference obviously doesn't mean having a 100% P preference, so there is a bit of wiggle room even with some unreliable questions, as most people with a P preference will answer yes to at least some of the J questions.