budget & time steps

I am running quasi steady-state model with rivers simulated as 1st kind with constraints (effectively high conductive drains). I ran the model for the same period of time (36500d) at various time steps:
-3650d
-365d
-36d
Looking at budget results, the 1st kind discharge rates stabilize at: ~6.4e4 m3/d (time step 3650d), ~1.5e4 m3/d (time step 365d) and ~4.3e3 m3/d (time step 36d) - which is quite a big difference.
Did anybody run into similar problems before?

As a comparison, it might be good to run the model with automatic time stepping, if run time allows.

I did that too - the time step icreases to about 4000d during the run and results are ~1.5e4 m3/d. I also ran the model with 1d time step which resulted in stable inflow rates of about 2.5e3 m3/d.

I am using Feflow 5.404, phreatic layer type, fairly detailed mesh.

Any suggestions of how to sort it out will be greatly appreciated.

Does the river have transient properties, or is the water level time-constant? If you're still having trouble, pls. send an email to our tech support (support@dhi-wasy.de), if possible with the model file (up to 8 MB). If the file is larger, our support team will tell you where you can upload the file.