Originally for training, there were used much more documents (20-90). To reduce the project size, most documents were deleted from the batches. However, even after retraining, the Samples matched column shows much more documents than the Samples count column. Does it mean that documents were not deleted properly and are still used somewhere in the project and occupy the space?
This is some statistics bug and it will be fixed in some future updates of FlexiCapture 12. Deleted documents from the training batches are not stored anywhere, so they do not occupy space.