Original User: gismare
Thank you all for your help.
I finally overcame the problem....
Indeed, the raster cell sizes for the kriged surface and the extracted raster were not matching up, even when choosing 'same as layer' (i.e. the input layer).
So, prior to kriging, I manually specified a cell size value (0.005 in this case) and did the same before extraction by mask. This gave a reasonable resolution around the peripheries, and for class boundaries.