The BC87 magnetotelluric data exhibit distortions due to three-dimensional structures at almost all scale sizes from very local to regional. Previous work has shown that these distortions can sometimes, but not always, be described by an electric field galvanic model and hence be removed using a tensor decomposition approach. This work describes re-processing of the original time series, using robust approaches to control leverage with jackknife error estimates of the decomposed regional impedanes, and extension of the analysis to include galvanic distortion of the magnetic field. The importance of the magnetic effects for many of the BC87 sites is shown However, even a combined electric and magnetic field galvanic distortion approach fails for quite a few of the sites studied. This is pervasive for periods shorter than 10~s, suggesting local 3D inductive effects, but model inadequacy at longer periods may suggest a breakdown of the model assumptions, particularly the requirement that the regional electric field be both uniform across the distorting body and comparable to that at the observation point.