I'd suggest you either find a calibrator to connect your meter to, or
make something you can use as a transfer standard. Maxim, for example,
has some reasonably low cost voltage references with good stability
(very few ppm/C). Then calibrate the transfer standard against some
known in-calibration volt meter or calibrator, at a known temperature,
and use it. I'd think in Scottsdale you could find a calibration
service that would do the job for you, or maybe you know someone who
has access to an accurate voltmeter. Of course, you should really do
it for several ranges. You can maybe calibrate a voltage divider
yourself, but it would save a lot of trouble if you could just do it
directly. If you don't know anyone who works with more accurate
instruements regularly and don't want to pay for a calibration, perhaps
you could pay a visit to the EE or physics department in Tempe and use
one of their instruments.

I've thought of making a secondary standard myself--I have several of
the reference boards from HP3455 DMMs that have single specific things
wrong with them, but combining them I have at least one good voltage
reference, and one good DC 10:1 divider.

Don't EVER expect to get to voltage measurements with accuracy or even
repeatability like you can with frequency--you would have to go to
extreme measures to avoid thermal EMFs, for example, and even with some
care they can give you part per million errors at low voltages. But
surely it's rare that you'd need that sort of resolution, let alone
accuracy, no?

Cheers,
Tom