Hi Les,
The time stamps in SAM are in local standard time, so 4:00 am local standard time in July would be equivalent to 5:00 local daylight time.
The results you see are a limitation of working with hourly solar resource data. The kilowatt value for each hour is the array output at the mid-point of the hour. In a real system, as the sun goes down, the array power would gradually decrease over the hour, but an hourly model cannot capture that effect.
Here are results from a simulation using 5-minute time steps for a 5 kW system Phoenix, Arizona that show the array output tapering off more than the results from the 60-minute time steps. The output for both subarrays stops at the same time when the sun angle indicates the sun is below the horizon, but the output of the east facing subarray is much less than that of the west-facing array.
I used the "Solar Resource Interpolation" macro (Click the Macros button under "Simulate") to create a weather file with 5-minute weather synthesized from 60-minute data. The results from the 5-minute simulation are not more accurate than the 60-minute results, but they do show how SAM's PV model responds to decreasing solar resource at the end of the day.
Best regards,
Paul.