The problem I find mecha, see the links on the post of slburris,
https://www.eevblog.com/forum/index.php?topic=533.msg7695#msg7695is that on that thread and on this link, Uni-T overstates its accuracy. If they quote their statistics as you say, that is still misleading. At least Fluke, Agilent, HP, B&K, and all other reputable makers I know either understate their specs not exaggerate them, so you know what you are getting. If it was calbrated properly but drifted, or they are quoting statistics improperly, its still not excusable. Further, those are new machines, not years old from calibration date.
But you are right, real accuracy, that not only is spot on but stable over time, is not cheap, but I can say with great confidence of you buy a known reputable make like Fluke or Agilent it will be what the spec sheet says it is; many really old Fluke 80 series DMM are still in spec,
decades since it was last calibrated [ I have 4 of them].
Accuracy is important when measuring electrical phenomena related to natural events: physiology, chemistry, etc., because in many instances the thresholds are truly finite, not relative. The cutoff voltage for NiMH cells is ~ 1.0 Vdc and to go lower risks damage to the cell. OTAH, deep discharging them close to this is a good way to prolong their service life, so the margin for error for good performance is narrow.
If your DUT provides 1.01V, meter should be 10x more accurate or measure to 0.001 or 0.1% to resolve 0.01V unambiguously. For example at 0.5% 1.01V becomes 1.005-1.015 and is more uncertain compared to 0.1% which provides 1.011-1.008.
...around 0.4%. ..to question the claim of 0.025% basic accuracy and 0.05% accuracy on the ranges. My expectation was that it would be with 0.1% accuracy at least.
...build a reference for the voltages of 1.9V and 19V easily enough but for 190V and 700V I am getting into expensive territory.
your concern about accuracy means you are embracing expensive territory. 0.4% at 19V is what? ±0.076V so your dmm read 18.924V? for that big number i usually just round off to 19V. at most electronics say 5V, err = ±0.02V 3.3V err = ±0.0132V
as saturation said. in addition, it maybe they are not lying, its just their interpretation on how to get the claimed figure. they maybe choosing the max accuracy dmm (instead of averaged or mod) from their testing and they probably have large standard deviation, meaning you get more chance of getting the lemon (out of spec) from your purchase.
i never characterize my UT71A properly, since its reading very close/comply to my other $20 DMM, so i'm happy. i dont really care much if its .1 out in 1st decimal, and i believe i'm blessed if the error is in 2nd decimal place. because i know... accuracy % = reciprocal $