I stated very clearly (somewhere ) that those references were only accurate enough to calibrate 3.5 and 4.5 digit DMM's. Once you get into 5.5 digit territory all bets are off.
Your ears burning when you should be sleeping
Have a glance at what I wrote re input impedance and check your references against low and high using the Siglent as you are effectively using them as transfer standards you did mention here somewhere you were getting discrepancy and that could be part of it between meters
Can't sleep tonight because of incident I mentioned earlier with my lady but I did find out from a co-worker that she was released from the ER and is home. I know her personality and she didn't want to talk with anyone and just go home and recover from her injuries. I'm sure she'll call me later today.
Anyway, I did notice a loading effect if you connected more than one DMM at a time to the references. So it's imperative to verify the output first with a known "good" DMM, then move the output to the DMM to be calibrated. Then go back and verify.
I have 2 4.5 digit DMM's that I trust as already accurate because I've never molested them and they match each other consistently. So those two are sort of a "reference". The Siglent obviously plays a role here too because it's new and it backs up my 4.5 digit references.
The arrangement ain't perfect and I don't claim it to be. But for a hobbyist like me is plenty good enough.
I also keep documented calibration records on how each DMM measures against my references. This is how I discovered that a Fluke 8600A was having drift issues and required repair.