Face detection seems to perform differently if 1 photo detected than multiple

[scope]

Hi,

Im playing around with the Face/Person detection feature of iMatch, and Im trying to understand something..

If I highlight more than 1 photo and run the face detection (with the "ignore images with existing face annotations" unchecked) it detects faces fine, assigning persons to some of them, however a number of them stay '?', or the wrong person is assigned to the face. 

But, if I focus on 1 of the photos and run the face detection (again, selecting "Remove existing face annotations"), for this 1 photo the person detection performs much better, assigning the correct persons to most of the faces.

Its as if the detection algorithm used is different?

Just for clarity, all people in the photos have been created, faces assigned to the people and each person has a some faces set as "train". 
I have tried different "Image Size" option when selecting multiple photos.

[Mario]

There should be no difference. The inner processing is always the same, whether one image or multiple images are selected.
In a quick test I could not reproduce this and this was never before reported.

I think what happens is:

IMatch detects faces and assigns persons in your first attempt with the multiple images.
It may not find a match for the first file, but it might find matches for some/all of the other files.
If a match is found and the person has less than 10 trained faces, IMatch uses the matched face to train the person.

If you now run detection again for the image where no person was found in the first attempt, the additional training may allow IMatch to match the face.