The first committed version of resampleDH.m includes a call to interp2 which is performed using the unmodified input array. This implies that the input data is not wrapped around across the -180 and +180 meridian.
Possible fix: pad input data with a copies of the input array.
Problems with this: the wrapped padding data should start/end at -180 and +180 (including room for step size), at the moment there is no check for this.