To make any sense of the distances, use projected data, that is an implicit assumption of distance measures since 'X' degrees is a pretty useless measure of distance without knowing location on a spherical body (eg think of the 1 degree 'distance' at the pole vs the equator. the results are indeed equal, but do not translate to euclidean space without transformation)
EDIT
If you are working with large interpoint spacing and you feel a geodesic calculation might yield a better result, then the function could be repurposed to do so. However, if it is relative distances within a group that is important then an appropriate projection would probably suffice. If memory serves Albers is often used to handle continental scale projections, particularly those with a large EW extent... Lambert conformal conic in canada. If the extent is largely NS and the points fall with the same UTM zone, then that would probably suffice (EW extent about 6 degrees give or take about 1.5 degrees on either side). Others could weigh in if they have comparative data to help you make a decision.