Your basic premise is that you need calibration-grade sources to determine that at meter is out of spec. This is false. You need calibration grade sources to properly calibrate the meter, and possibly to verify that it is in spec. If your Siglent DMM is reading 2X actual values, I can demonstrate that with a Simpson 260 or a Harbor Freight freebie.
That would be true for a CALIBRATED Simpson260 or a Harbor Freight freebie. And then it would be only possible to do so if error was larger than what the guardbands are for Simpson.
Answer from Siglent is correct but not written in language easily understandable for those that are not clear on all the concepts.
Let's break it down:
1. Factory specification for meter is
after 90 minutes of warmup (1 hour 30 minutes). Anything before that is IRRELEVANT to specification verification. By the way, that is somewhat similar for all long scale meters. I believe someone mentioned than new Keithely meters are bit faster in warmup, less than 45 min to good accuracy. My old Yokogawa takes 2-3 hours to settle (no forced heating), Rigol DM3068 60-90 minutes, depending on temp in a lab.
2. They tested meter with Fluke calibrator and factory calibration fixture and it was
verified in spec. That alone
does mean it is passing verification.
3. Then they went to try and demonstrate how people can get confused by using equipment that is not good enough to test accuracy. OP tested with bench PSU, and 20+years old Fluke 187 that was never calibrated and a noname cheapo multimeter. Why would you proclaim brand new 6.5 digit meter is wrong when tested against 20 year old handheld instrument of inferior specification, whose temperature coefficient in 1°C contributes more than TWICE complete accuracy specification the instrument ??
Let me repeat: 1°C of temperature difference for Fluke contributes to error
TWICE as much as full specified error for that range. 1°C of difference and your error in that range is not anymore 0.03% but a
full 0.08%.. Few more degrees and all bets are off..
So they went and tried to demonstrate how you CANNOT do that with inferior instruments, but that got misunderstood. They didn't try to demonstrate how you can do performance verification with B2901A, but opposite, how you cannot do that, even with such an expensive SMU, because even THAT is inferior to what is needed to verify 6.5 digit meter.. He was probably confused how OP doesn't understand those basic facts.
All this is just a demonstration what kind of bias exists out there. These people are not amateurs. They know their stuff.
So let's do some math for worst case scenarios..
Quick calc for Fluke 187 (providing it is FULL in spec):
At calibrated temperature What can it really be:
Accuracy Measured Min Max P-P error
0,0300% + 0,003 . 30,04100E+0 . 30,02899E+0 30,05301E+0 24,02460E-3
0,0300% + 0,003 . 10,01700E+0 . 10,01099E+0 10,02301E+0 12,01020E-3
0,0300% + 0,003 . 59,99000E+0 . 59,96900E+0 60,01100E+0 41,99400E-3
At 1°C+calibrated temperature
Accuracy Measured Min Max P-P error
0,0800% + 0,003 . 30,04100E+0 . 30,01397E+0 30,06803E+0 54,06560E-3
0,0800% + 0,003 . 10,01700E+0 . 10,00599E+0 10,02801E+0 22,02720E-3
0,0800% + 0,003 . 59,99000E+0 . 59,93901E+0 60,04099E+0 101,98400E-3
So WITHOUT even adding SDM3065X specification, what is shown is
included in error specification of THE Fluke 187. And by that I mean that nothing is wrong with it, it just how accurate(inaccurate) it is. Which, by the way, is excellent for 60000 count handheld instrument that old.
So OP's findings are just fine. But conclusions were wrong. Errors made:
1. 90 minutes to warm up.
2. Not calculating what specifications really mean
3. Cognitive bias that 20 years old inferior specification Fluke must be "more right" that Siglent's modern, new 6.5 digit meter...
4. Not being suspicious and going back to it, until all the math is right.
Those are all human failings. Nothing unusual to make mistakes. But they have to be opportunities for learning. And if we put in effort, it can be opportunity for all of us to learn new things.. It makes us better..