Thanks for sharing your inputs. I just requested access to that GDrive folder.
SDR results depend on how different land cover classes are parameterized in the biophysical table, namely the values for cover-management (C) and support practice factors (P). Lower values will lead to less soil loss (USLE) and sediment export, while higher values will cause the model to predict greater USLE and export. To attain the results you may expect, the conservation infrastructure features you mentioned must be represented in the LULC input with corresponding C and P values reflected in the biophysical table. See this section of the User Guide for more information.
An important note about dams and the SDR model, is that once sediment reaches a stream or river, the model assumes it ends up at the catchment outlet, thus no in-stream processes are modeled. To deal with this, we recommend delineating watersheds/subwatersheds by inputting dams points as outlet features. See the Appendix for how to do so. The final paragraph of this section also discusses how to handle dams.
If you continue to need assistance, please let us know here and be sure to grant me access to the GDrive folder you shared. Please also be sure that the log file (.txt) is provided.