The gold standard in this case usually is a meter that has tighter specs due to frequent calibration and lower uncertainties used for it and a better tolerance 3458A (HFL or 002). They probably dont take the effort to characterize it additionally, just too much effort, the HFL will do. The standard 3458A calibration by Keysight, as per manual, is based on standards with 2ppm (10V) and 3pm (10k) uncertainty, which needs to be added to the spec uncertainties of the 3458A. This is not very tight. Uncertainties under 1ppm are possible without too much effort. Not sure how they applied the 3458A in your specific case, but in most cases (DCV...) the 5730 is the more relevant standard. Also, theer are different calibrations performed by Kesight, there is e.g. info here on eevblog. Also, there is a document by Fluke talking about how to calibrate the 3458A with a 5730 and what is critical and how to combine uncertainties re. a pass/fail. When you go through it you will see the limitations of this procedure. Your cal report will also give you some insights into what uncertainties they apply and what they did specifically. Might be interesting to see it here (if you have permision, and keep in mind the cal report also has certain limitations, e.g. reproduction only in its complete form). As per standard calibration procedure of the 3458A cal manual, the gold standard 3458A is used e.g. for AC comparision (SYNC versus ANA mode).