I think what you described are pieces of the puzzle that describe the story. We work through a series of tiered risk assessments, starting with the very most fundamental, controlled, in-lab types of studies, which give us direction as to what we should look for at the next level. Eventually you get out into the environment you're monitoring, which is the highest-tiered risk study you can possibly run.
Modelling is used to predict what you might find out there. Obviously, a model and real-world monitoring in a perfect world will match up. You'll find what you're looking for. A mesocosm study simulates the natural environment but tries to control the variables, which are numerous when you go out into a natural body such as a wetland or a stream running through an agricultural area.
In this case, all these studies were not considered in the assessment. They were deemed for whatever reason to be inadequate, although previously, I will say, they were used to grant the initial registrations.