QTR-Quadtone RIP

My workflow:
- create ink profile in QTR
- select curve and print 21 step tablet
- measure steps after dry-down using X-Rite 810
- enter measured values into QTR linearization array and regenerate curve

Problem:
Steps never achieve the ideal density for the measured dMax.
When I linearize using Photoshop transfer curves, I can get within
.01 for all steps, but I am often off by at least .05 - .10 in
multiple steps when I linearize using QTR. Multiple iterations through
the process don't improve matters; things often get worse. 

Can anyone shed some light on this? I assume I should be entering
*measured* values into the linearization array, and not *derired*
values. Too bad QTR doesn't present the interpolated curve and let
users edit it directly ...