Hi Lamm,
generally it is better to extend your model border to the natural borders / watersheds.
One possibility is for example, to build a large regional model covering your local model area of interest to define regional groundwater regime. This could be a quite coarse model with rough boundary conditions.
After that, you can extract a local model of your regional model and set the lateral model inflow as simulated from the regional model.
Mostly, this processing can be much better explained to clients, than assuming or guessing lateral inflow.
Best regards, Christian