Why were we so wrong in estimating field errors in SSC dipoles?
Generally there are 25-50 micron (1-2 mil) error in parts and construction. Therefore, allow this kind of positional error in each of several blocks of conductor (see picture below) and then sum the resultant field errors in an RMS sort of way.
Current Thinking (personal opinion)
The errors in parts do not necessarily translate to the error in field harmonics. The effect gets significantly reduced from averaging and symmetry considerations. For example consider how a systematic or random error in collar, wedge, cable, coil curing plays in a real magnet.
Movement in popular models: one red arrow
Symmetric model: 4 black arrows
Realistic model: some thing in between but closer to black arrows