Thanks for the thumbs-up and the new graphs, guys.
Cathar, some accuracy was lost in GPM-LPM conversions, eyeballing, and polynomial/power function best fit approximations. Therefore, I'm almost surprised that you only found one point in disagreement
(Your graph will inherently me more accurate than mine.)
As you have already said, it would be useful to have the raw data available somewhere on the site to avoid some of these issues.
Edit: Choice of pump seems quite determinant here
(My only quibble is with the 15w of heat produced by the D4. Is this at 12v? )