I am also having a similar problem (see pictures attached)
I have set different 3r kind BC (Cauchy) in a 3D model (on the top slice and with all the nodes in the river's vicinity selected). All the condition are suposed to be an exit of the model for the water: so the in-transfer rate is set equal to 0, and the out transfer is set to a positive value.
What I find in the rate budget is that there is a very high value for both in and out rate on the Cauchy BCs bar. The difference between these values is sometimes an exit rate that I could interpret as the real water leaving the model in that area, but many other times it just doesn't make any sense... I really need to understand how much water is leaving the model to validate it.
Also, by checking the head distribution it doesn't seem that there is any place where the water is entering the model through the BC.
Thank you very much,