Azizi:

This is what is known as "cosine error"... it comes from not gathering points normal to the surface... you probably have no repeatability either... the best way to overcome this error is to get the distance from the average center "datum" to the center of the probe... this can be done by over-riding the probe radius and forcing it to be -0-... once you have the distances to the center of the ball mathematically add back in the probe radius... don't be discouraged... this happens with manually gathered data on things that are not "approachable"... you are probably not at 15 deg increments either... you can report individual calculations as polar radians or average any group of them for a radius... remember 3 points make a perfect circle... any more, be sure to evaluate your form...
http://www.qualitytechnologies.com