I never said your meter was definitely not high.
I said nothing you posted _yet_ supports that conclusion, it is consistent with an in spec meter and a meter that is high by that amount and perhaps (depending on your meters published specs) one that is low by some percent.
So you're already moving the goalposts from your original statement: "All your measurements are consistent with your meter being spot on spec."
What? this is exactly the same statement.
measurements can be consistent with more than one thing that is why I said you cannot conclude anything about your meter yet. They are consistent with the meter being on spec because the measured values are all within the tolerance of the wing shing TL431A part you specified so those measurements can be spot on. They are also consistent with it being off the nominal value by 0.5% because the TL431A is a 1% part so could be that far in either direction.
That you cannot in general average a bunch of 1% parts to get a more accurate reading is also well known and empirically tested in daves video.
Now, you may have 0.3% parts, or may not. But when you say wing shing TL431A I can only assume wing shings data on their TL431A part is accurate rather than suppositions from an aliexpress listing, in any case, wing shings official specs are all that matters for making engineering decisions rather than a batch that happened to be 0.3%. Also, no matter what the alibaba listing says, I'd say wing shings datasheet trumps it.
If you bothered to read the full thread, it's not just the tl431a's that suggest my meter is high, the alkaline battery measurements and the SR43 battery measurements suggest that as well. Yet you seem to think that my >10yr old "Mastercraft" CAD$20 meter is a more reliable reference than a sample of 11 new (2015 mrf date) TL431As.
Look, this is metrology. specifications mean very precise things and the conclusions you can reach are well definied. As an engineer, you have to assume the worst. 1% means 1% out in the opposite direction. battery chemistry is so variable (the mercury cell being an unusual exception, my keithley electrometer uses one as its voltage standard, still spot on after >20 years) you cannot definitively make decisions based on it. Like you derate capacitor voltages because i may have the worst batch of the lot and my design might encounter some crazy transients.
In spec for a meter does not mean at its nominal value, it means within 0.1% 1% or whatever the meter guarentees. When a meter guarentees 1% accuracy, it doesn't mean they were only able to trim it within 1%, of course they could trim it to be spot on, on that day, in that lab. but if they know the parts have a 1% drift over temperature range or expected age, or that their calibration meter is only good for 1% then they rate it as 1% accurate so that is stays within spec during its lifetime. When you calibrate a 1% meter that is at 0.9% off, they don't twist the knob back to 0% because that is in spec.
I am not sure why you think these general engineering practices are somehow contrary to what you say. It does not _matter_ whether your meter is on or off for these statements to be true or not. That you cannot conclude _definitively_ (not probably) it is off by 0.5% with a 1% part when the measurements are within the 1% margin (even averaged, if they have not been characterized) is, well, pretty obvious.
Again, I have to go by manufacturer data sheets and part numbers as specified. Once you get your ref, you will be able to characterize your tl431s and then say definitively how far out your meter is (up to the accuracy of the ref5050)
I see another excuse in the making, "You mislead me to think that your meter was new..."
An excuse for what, your measurements _are_ consistent with an in spec meter for TL431As as speced by the manufacturer. They are also consistent with a 0.5% off amount. Our statements are compatible. It would surprise me somewhat less that your meter is out of spec since it is older, but that doesn't mean the original observation that you cannot conclude definitively yet is false.
There are many people reading this forum that have to build things that work without relying on probably getting 0.3% parts marked 1% time after time, or that like or need to be _certain_ of the tolerances of their measurements are finitely bounded by a known number.
On an engineering board, where we constantly have to take into account _worst case_ (but in spec) scenarios day in and day out, and run into horribly expensive or dangerous failures when someone doesn't take into account the possibility that your 1% parts might actually all be 1% out the wrong way or other allowed things then it is natural for these concerns to come up.
I was trying to be helpful by bringing attention to a conclusion that did not yet have enough evidence, the same as I do for design reviews where I am paid to notice these things before they become issues, or become the subject of an audit on who dropped the ball on specing parts properly. Even if you don't seem to get what my statement actually means, and why it is consistent with yours, Consider it good advice for the community and an incidental benefit of this thread.