As expected, the northern half is too smooth, but the major trends are captured in both examples. The general shapes of the major ridge and the western dome features continue into the northern half.
It is possible that this model with two different data densities needs two different regularization parameters, . In Figure 2, the southern half is clearly being blurred by the regularization fitting goal. The used in this case was the that created a reasonable image in the northern half. In the southern half, the best to use would actually be 0, because on this grid there is no missing data in the southern half.