Hi Brian,
Here are some more thoughts on this.
The following table shows inputs and results for five systems in the attached .sam file. I started with the file you attached above, but removed System 6 because it was the same as System 4 and added a System 7 with a 1500 V inverter, and a System 7 with the same inverter but modified temperature derate curves.
The inverters for Systems 5 and 3 have lower maximum MPPT voltage ratings than the others, so SAM's array autosizing on the System Design page chose shorter string lengths (15 and 18 modules, respectively). These shorter string lengths result in lower operating voltages. On the Statistics tab of the Results page, you can see that maximum ambient temperature for this weather file is 34 degrees Celsius, and the maximum operating voltage for these systems is below 800 VDC. Given the default inverter thermal derate curves, these inverters experience no inverter thermal derate losses.
System 4 uses an inverter with 1200 VDC maximum MPPT voltage, so SAM chose a string length of 23 modules, which results in a higher maximum operating voltage of 946 VDC. On the Results page PDF / CDF tab, I created a histogram that shows that the array often operates at voltages above 800 VDC, so it makes sense that it would result in more losses due to the temperature derate curves. These power losses reduce the inverter power so that it does not exceed its nameplate rated power, so the system experiences no power limiting losses.
System 7 has the highest DC/AC ratio and inverter nominal voltage, so we would expect it to have the highest power limiting losses and thermal losses. Its maximum operating voltage is 1,111 VDC, but we see thermal derate losses of only 101,441 kWh/yr. Because it is hard to visualize how the inverter thermal derate curves extrapolate from 800 V to 1500 V, I decided to see if the inverter datasheet has any information about response to temperature. The datasheet has a category called "Operating ambient temperature range" of "-30 to 60 degrees C (> 50 degrees C derating)". To model that, I created the system-7 mod case, which is a copy of the system-7 case with a thermal derate curve that matches that description:
I hope that helps. Let us know if you have other questions.
Best regards,
Paul.