The (b-y) and c1 indices are smaller for a peculiar than for a normal star. So in the c1 vs. (b-y) diagram used by Crawford (1978) to establish the standard relation for early normal stars, the peculiar stars are often below this standard relation and yield negative reddenings. The criterion by Philip et al. (1976) to separate peculiar stars is based on such negative reddenings. So, the excess is often underestimated if the peculiar stars are treated as normal stars.
Figure 10: Difference in colour excess computed from Johnson photometry
by the Q-method and from 
photometry for normal
stars belonging to the early region
Figure 11: Correlation between the difference of colour excess due to
peculiarity for CP2 stars belonging to the early region
Since Johnson filters are wider than filters, the Johnson
colour indices are not greatly affected by the peculiarity. So,
when treating the peculiar stars as if they were normal stars,
the colour excess computed from Johnson photometry is more realistic
than when computing from Strömgren-Crawford photometry. We compared
the colour excesses computed from both photometries for CP2 (284)
and normal (844) stars of the early region.
To compute E(B-V), we applied the Q-method using E(U-B)/E(B-V)=0.72
(Crawford & Mandwewala 1976) and the luminosity class V calibration by
Schmidt-Kaler (1982) for the relation (U-B) vs. (B-V). The E(B-V)
was converted to E(b-y)Q by the factor 3/4.27. The E(b-y) was computed
from Crawford's (1978) calibration for normal main-sequence stars.
Differences 
in the sense E(b-y)Q-E(b-y) are plotted
vs. [u-b] index in Fig. 10 (click here) for the normal stars.
This reflects the intrinsic differences of the two methods
when computing reddening, i.e. the differences between the standard
calibrations
used. However, the of the CP2 stars includes the difference
due to the calibrations and the difference induced by the peculiarity.
For normal stars can be fitted by:
Subtracting this difference from the of CP2 stars, the new
will only be due to the peculiarity of the star. The
correlation with 
is shown in Fig. 11 (click here) and is fitted by:
Maitzen & Vogt (1983) performed an analysis of the colour excess,
computed from Crawford's calibration, of CP2 stars located in an almost
reddening-free sky region. They deduced a moderate correlation
between the negative reddening values and the peculiarity index 
.
Our correlation is stronger than theirs.
With this 
the number of negative excesses in our sample
is reduced considerably.
Since without a spectrum we cannot know a priori
if a star is peculiar, the best solution is to use the parameter
given by Eqs. (1) and (2) and to correct the excess a posteriori
with the preceding expression if the star is classified as CP2.