leonardoa I do remember we already had discussions with you about recognition accuracy and that sometimes it makes mistakes, and this is just the way it is. Here is just another example of it. In our own testing it also comes up with results like that from time to time, but since this is around of 1% of its suggestions according to our experience, we consider it as totally within the norm. This is always a balance of multiple factors, and we consider this is very small price to pay for the fact that algorithm is capable to do clusterization on hundreds of thousands of portraits within a fraction of a second.
So unless this is something program provides you majority of the times, then I would not consider this as a bug.
