> Parapsychology is physically impossible, and the evidentiary standards in physics are much higher, so we have much more confidence in our physics results than in these experiments
Conflicting results don't mean one set is impossible (cf., the long-standing apparent conflict between QM and relativity within physics). Apparently conflicting results without a methodological error in either imply that the explanatory model that appears to be supported by at least one of the results (if not both) is, while useful within its own domain, in some way incorrect.
The whole idea that the models validated by scientific experimentation are binarily true of false is, well, missing the point badly. While over time we hope they approach truth with them, what they are is useful (that is, they have predictive power) to a greater or lesser extent. And quite often the models with the greatest predictive power in two different domains conflict when either or both are extend outside of their own domain.
EDIT: The real problem with parapsychology is that there's little in the way of explanatory models being tested anywhere in the field. There's a lot of hypotheses without models and some experiments testing them, which (concerns about methodology aside, for the moment) might raise interesting questions and serve as inspiration for developing and then testing theoretical models to explain the effects, but very little has been done there -- which makes "parapsychology" more a collection of potentially unexplained phenomena more than a branch of science that provides an explanatory model for some set of phenomena.
Which is very different from most of the social sciences.