I'm still cross linking loneos.phot of Brian Skiff's against UCAC3 via VizieR, which always takes forever for my setup for some reason, this crosslinking, but a preliminary look for 3500 ish matches of loneos.phot V against UCAC3 model fit "f" magnitude and 2MASS J and Ks shows the following :-

There are some UCAC3 magnitudes of less than ten mags, down to 7 in places, but these are few relatively and reliable V magnitues are available elsewhere. So the lower limit was set at 10.

There's a dogleg, a small break, in gradient at model fit magnitude 14.5.

This is a break, often people when they see breaks will do a polynomial fit. This isn't nonlinearity, this is a break, with two separate linear regimes. The faint stuff increases in scatter pretty quickly too.

There are not a few outliers, so any particular magnitude determination is not guaranteed, some sort of ensemble approach is best/safest.

Thus with these caveats aplenty for the colour range J-Ks 0.0 to 1.0 (as sourced via UCAC3) and the magnitude range UCAC3 model fit mag (here called "f") of 10 to 14.5, the following results.

V = f - 0.095(f) + 0.6(J-Ks) + 0.92 standard deviation 0.14

For ~3500 objects of that colour and magnitude range that formula gave a standard deviation (1 sigma, ie two-thirds-ish chance of lying within +/- that range about the central result) of 0.14 when calculated V was subtracted from loneos.phot listed V with mean on that same list of objects of -0.006 and median value -0.004.

2 sigma, ie around 95% of the time, is then +/- 0.28 accuracy, which is just about the 0.3 magnitude full range mentioned for UCAC2 in the past.

Notice there is _both_ a linear colour trend and a linear brightness trend involved. The fainter the magnitude the more faint f magnitude is relative V. A plot of the aperture mag superposed on the model fit mag shows there's little difference between the two in terms of value and distribution, so I expect similar results from that.

Still not an ideal photometry source, but probably a better source than any photographic magnitudes or system especially as those are usually saturated at 10 to 14.5th mag, and things like Tycho2 have a useful lower limit of about 10.5th mag. However, single object determinations are not guaranteed due to outliers, although quality flag usage may correct for that at times (ie by rejecting poor candidates). Using them as an ensemble source might well be better than using photographic plate derived magnitudes.

CMC14 derived V (V = r' + 0.6(J-Ks) - 0.03) has a larger magnitude range it works under, going down to about mag 15.5 ish, for a similar J-Ks colour range, but it does not have all sky coverage.

Summary : if you've got to use something, and really need to with no alternatives, UCAC3 derived V mag for UCAC3 mags between 10 and 14.5 and J-Ks colour between 0 and 1 will be better than V mags derived from GSC or similar, but there are sufficient outliers for no particular single object determination to be entirely safe.

Remember, UCAC is an astrometric catalogue, and its developers and authors have never made any claims upon it being a photometric one, although they have strived to provide the best possible magnitude estimates feasible.

Cheers

John