The RHIC dipole coil cross-section was optimized for small b2 at design field and small b4 at injection (yoke was optimized for small variations in between). The final phase of dipole production showed that on the average both were much smaller than the geometric tolerances (25 micron or 1 mil) in parts. Even a 10 micron systematic error in the critical wedge would have generated larger systematic harmonic errors than measured.
The RHIC magnet design philosophy was based on a flexible approach where a mid-course correction in the manufacturing could be easily applied without disrupting the production. This approach accommodated geometric errors in individual parts, kept production line moving smoothly and made magnets with average field errors less than the geometric tolerances in parts.
RMS errors (shown earlier) were also much smaller than previously thought possible. RHIC dipoles use phenolic RX630 spacers between the coil and iron. This is a critical component which defines coil geometry and hence influences the geometric errors in field harmonics. This component had part to part variation of 2-3 mils instead of 1 mil. However, the RMS errors in RHIC magnets (generated from this and other parts) were much smaller. Explanation: the field errors are smaller than the corresponding mechanical errors in parts thanks to averaging (if the quantity of those components is large) and symmetry effects (if the components are used in a symmetric fashion).
Conclusion: Both systematic and random geometric field errors in magnets are much smaller than the geometric tolerances in parts.