The Warp operation failed because of an incorrect polynomial transformation. Polynomial transformation uses a set of control points to determine the mapping of input coordinates to output coordinates. When the control points are correlated or not well distributed, the warp result is often unexpected or has an exaggerated Root Mean Square error. The polynomial transformation is incorrect under such conditions.
When picking the control points for Warp, make sure the points are evenly distributed across the image. This way, the chance of correlation among the control points is reduced and the RMS error would be well distributed among the control points. Also, make sure enough control points have been collected for the transformation. The number of noncorrelated control points required for a first order polynomial transformation is 3, 6 for a second order, and 10 for a third order.