A few suggestions:

• Maybe isolate the investigation to 1x zone (with 1x window) in your model, e.g. plot solar transmission vs zone cooling vs heating vs (variable) ventilation, even looking at autosizing of HVAC equipment. Further isolate the tests to summer vs winter design days vs off-season (less than 1% difference in results).
• I work in a cold climate, with significant peak cooling. I do expect increased kWh with lower SHGC in many circumstances.
• Many variables may be at play, e.g. dominant facade orientations, %FWR, internal loads/scheduling, etc.

First guess is that E+'s Simple Window Model may not be the issue here. Based on the stated glazing properties, the windows should fall right in the middle of option 13 (curves FGHI). Sharp changes may occur when the model switches between curve sets, but not in this case IMHO.

For what it's worth, I ran the same window types in E+ 22.2 for a residential home in Denver, and the results were smooth and as expected. So it might be unique to your model. Are you using the latest E+ release? Can you provide IDF files?

@Denis Bourgeois thanks for your suggestions. I agree with you that the SimpleGlazingSystem may not be the issue in this case. I will examine the model following your suggestions.

@shorowit I am still using E+ 9.3 but I can get reasonable results in the other models. Maybe it might be unique to this model. Thank you for your help.