Sometimes the effort data points on my Life Cycle Sensitivity to PI graph have values that do not always increase (or decrease); e.g., there are small dips in the data.
Remember that the time and effort shown on these graphs are life cycle values. The most likely cause of slight variations in the expected behavior of the data is caused by variation in the effort for phase 4. Staffing projections for phase 4 are dependent on the staffing rate at the end of phase 3. Therefore, differences in the final phase 3 staffing rate can result in variations in the total effort for phase 4. You can check to see if this is the case by making the solutions in question the current solution, then comparing the phase 4 effort in the solution panel of the staffing view.