1. Diffuse sources: I typically use this input to represent groundwater baseflow entering the stream across a reach. As an example, if an upstream gage reads 10 cms, and a downstream gage reads 20 cms and there are no point sources or tributaries, it is like that 10 cms is entering the stream via diffuse groundwater inflow.
2. Importance: depends on the system! If you are modeling a river under a critically low flow period, the diffuse flow may be small. If you are modeling a losing stream where water leaves the river into the sediment, it may be important to capture that loss with a negative diffuse flow. I have worked on many modeling applications that have used diffuse flow and have NOT used it, so it really depends on the hydrology of the system you are trying to capture.
3. Data Inputs: if you you don't have any data, you can refer to literature for assumptions or leave inputs related to water chemistry blank potentially. Sometimes nearby groundwater wells will give you an idea of typical water chemistry of the groundwater, or literature about reference values typical of groundwater in your region based on geology.
Hope that helps a little!
-Hillary Yonce