Four possible causes for the failure of the Mtera-Kidatu Reservoir System within the Rufiji River Basin in Tanzania in the early 1990s were investigated. These were sudden decrease in inflows, sudden increase in losses, sudden increase in hydropower generation, and unnecessary spills; or a combination of these, and it was found out that unaccounted for and unnecessary spillage was the main cause. This paper proposes that consideration of the flows that are generated within the intervening catchment (i.e. catchment between Mtera and Kidatu) and the operational policy that maximum power is produced at Kidatu most of the time must be the core in the management of the reservoir system. If this was the case in the past then the Mtera Reservoir should not have gone dry in the 1991-1994 period. The validity of this assertion was tested with the TALSIM 2.0 model and an efficiency of 95% was achieved, indicating a very good correlation with the investigative techniques employed in this study.