rc;

Concerning your more than needed data points, have you tried reducing 
the number of points before running a spline?

Case in point, in contouring 3D data, every once in a while I 
encounter data that is conflicted. In other words, very close x/y 
data with anomalous z values. In such cases, I often use a routine to 
figure out if one, or both, of the z values are out-of-range for the 
data set. If one is, then I throw out the bad one and use the good 
one. If both are bad, then I throw both out. But, if both are within 
tolerance for a global range but are anomalous for a local range, 
then I average the x, y, and z values and replace them with the point.

Likewise, to solve your knot problem, you could write a routine that 
would check for points "too" close together and then average the 
point before running a spline.

Just a thought.

tedd