After looking at this more closely, I seem to have found a solution but not really an answer to my original question. My personal conclusion is that the Area and Volume tools using the methods in my original post are flawed - a strong statement indeed, and I'd be grateful for others opinion and explanations to the contrary.
In the end I re-calculated all my volumes using the Cut and Fill tool using a zero value raster as the plane. The results are consistent, understandable and ultimately make sense. I would still like to know why the volume and area tools do not give correct values and would welcome any comment from esri on whether this is an issue that needs resolving and if not - why not? As per the link in my OP, the issue seems to have remained for over 6 years!
My recommendation is; be very careful when relying on the area and volume values reported by the Surface Volume tool and Area and Volume Statistics, especially if you have irregular, relatively small and coarse grid rasters. The slightly more time consuming but more accurate method to get area and volume stats is to create a single value raster at a plane value you would have used in the Surface Volume tool and then use the Cut and Fill tool instead using this raster and the layer in question.
