HEC 22 loss method


I am writing regarding some results using the HEC 22 loss method. We are working on a sewer model using the explicit calculation engine and HEC 22 loss method by request of our client, so this cannot be modified. We calibrate the model using field data basing on flow, depth and velocity. We have found cases in which depth results show an unexpected jump due to HEC22 losses. See the following screenshot:


We wanted to ask you about these jumps because it does not seem to conform to real behaviour, and there is nothing similar on any of our meters.

In short, why do these jumps appear? Is there a real explanation?