Fig. 2 shows our visible-region mean abundances versus the SD93 UV-region abundances. The stars exhibit a very good match between the UV abundances and our visible-region abundances, with an apparently random scatter on the order of 0.1-0.2dex. Only three stars show larger discrepancies, all of which are binary systems in which both components' spectra are visible.
We note that SD93 made no allowance for binarity in their results, on
the assumption that the secondary's UV flux would be unimportant,
except for a few rare cases (they cite HR 1800, HR 4072,
CrB and 
Lup as being possible exceptions). In the
current visible-region study, we have accounted for dilution by
scaling the observed 
for the Mn lines in accordance with
the light ratio deduced for each binary system (see Observations ).
If binarity were ignored, the true strength of the Mn lines would be
underestimated and thus the derived abundance would be too low. Indeed
this is exactly what we find for the UV results when compared with
visible-region results. The worst offender is the double-lined binary
Lup. In this system the UV lines should be significantly diluted by
the companion star, leading to an underestimate of the UV Mn abundance,
exactly as found. The other stars showing larger discrepancies from our
mean results (HR 1800, 
CrB,
HR 4072) are double-lined binaries, and show
greater Mn abundances in the visible than the UV, again as expected. The
other binary systems appear to have negligible UV contribution from the
secondary star and their treatment as single by SD93 would appear
justified in that context.
Figure 2: Mn abundances from visible and UV spectra. The solid
line is that of equality. The dotted lines indicate a 
0.2dex deviation from equality. All three stars substantially
exceeding these limits are binary systems (shown with arrows).