How can I reduce the size of the modal data imported in a hybrid model?

The model will always contain enough data to solve, and give basic results. For models with hybrid area junctions, nodal data for all wetted faces will be generated/recovered which results in much data.