Ah, sorry I can see I was not so clear now... I missed a small fact 
2.5million in the full data set, BUT only 200 rows in the data preview.
I would assume that HANA peals off the top 200, and then applies the calculation.
If it is applying the calculation first, then... well, oops!