How many points in a trajectory is reasonable

edit

I am using a Jaco2 arm (DOF) on Moveit/indigo/14.04. I have a small program written in Python (using moveit_commander) that moves the arm in a simple box like movement (left, down, right, up) with the typical movement of about 3" in the horizontal (X) range and 8" in the vertical (Z) range. The Y range does not change.

I am finding that Trac_ik is generating between 10 and 50 points for virtually the same movement distance. The number of points generated seems pretty random although it seems that I either get 10 points or something above 30. I have the solve type set to Manipulate2 (also tried Speed, Distance with no change in behavior). The 10 point behavior generates nice movement that has little error while points above 20 or so generate odd, extreme arm movements that seem to end somewhat far (6+") from the goal. When that happens I replan to the original destination but it may still take 3-4 attempts to get to the correct goal.

I have tried filtering out plans with too many points but it can take 10 or more attempts to get a reasonable set of points and even that number of attempts may not yield a good count.

Is this normal behavior or am I doing something really wrong