I did think carefully about the inner cone. If it had a flat 2 mm wide on the end it would give a theoretical error of + 0.03mm on a 30mm ID tube, less on a bigger tube. I machined the cone to a point which should eliminate this error to practically zero.
I thought miss-alignment of the cone and DTI would be a more likely cause of error but this would be negative and pretty darn small.
All tubes were tested along 3 lines at each end and aligned to give the minimum reading at each point.

I tried to be scientific about it and am confident that any errors are small.

Not sure what to do now but will contact Reynolds and see what they say.