Lower error spatial calibration than phantom registration

[ndehaan123]

I performed the fcal calibration of plus with elektromagnetic tracking. The probe of NDI Aurora is attached to my transducer and I perform a spatial calibration to determine the transformation matrix between the image and probe. I performed the calibration twice and for the first calibration I got the following errors: stylus calibraton: 1.01 mm, phantom registration: 4.13 mm and spatial calibration error: 0.579 mm. The second calibration I got the errors: stylus calibration: 1.00 mm, phantom registration: 2.86 mm and spatial calibration error: 0.733. The reference is attached to the phantom.

My question is how exactly the spatial calibration error is based on the error of the phantom registration. I dont understand how it is possible that I get a larger spatial calibration error when my phantom registration is more accuracte compared to the other calibration. What is exactly the influence of the phantom registration error to the determination of the spatial calibration?