Regarding ascertaining a two knot speed increase, how did you establish that? I'm not suggesting the increase is not real, it is just that there are so many variables. A slight change in temperature, density altitude, mixture, rpm, weight, bug splats, instrument interpretation, climb/descent, etc. can easily be more than two knots. I have found the process of speed data collection to be very, very difficult, requiring many flights, and even then I'm not sure. Kevin Horton may have some thoughts if he sees this thread.
An interesting comment on wing/fuselage intersection fairings - a number of years ago someone with a computational fluid dynamics software package digitized an RV6A, and tried numerous different configurations of wing root fairings. None resulted in a measurably speed increase on that aircraft. I recall vaguely that they understood why, but I don't recall the details.