To fix your problem, try extending your UZ ‘soil’ profile further downwards to cover the full range depths you expect for the water table: adjust both the ‘Soil Profile’ table and calculation cell ‘Vertical Discretization’ table for the relevant soil/material profiles. NB: The MIKE-SHE interface labels this as a ‘Soil Profile’, but it may need to extend beyond what is often defined as ‘soil’, into weathered rock layers that are frequently unsaturated.
You may also need to adjust the thickness (or top and bottom elevations) of your top SZ calculation layer so that the water table also lies within that top layer throughout the simulation. Getting these depths right for the full length of the simulation may be an iterative process if the groundwater table is predicted to move up and down by large amounts.