The microlensing candidates found in the first DUO season are presente
d here. Statistical tests show that these candidates stand well above
the noise distribution. The analysis of their colours shows in additio
n that they are not related with pulsating variable stars. We also con
clude that they are unlikely to be associated with cataclysmic variabl
es. Unless a new class of variable objects has been discovered here, m
icrolensing appears as the best explanation of their light curves. The
latitude distribution of these candidates is different from the distr
ibution of the other stars: it shows a large gradient which is well re
produced by a microlensing model. We conclude that this is a good sign
ature of microlensing. The large fraction of short duration events amo
ng these candidates is rather puzzling, and cannot be reproduced, with
in a standard microlensing model, without a large number of brown dwar
fs. However, a more sophisticated model including the amplification of
unresolved stars can well reproduce the observed rates without any br
own dwarfs. Consequently, we demonstrate with these data the great imp
ortance of modeling the amplification rates of unresolved stars. It is
also concluded that any satisfactory study of the mass function at th
e faint end, should include this kind of contribution.