The ways of FCP X rendering are still mysterious and unknown. Sometimes exporting without rendering first is quicker than rendering. Other times rendering first then exporting is faster than exporting without rendering. It’s a mystery. My guess is that depending on the codec used and the codec exported to, it will sometimes utilize the render files, while other times it deems them to not be appropriate and rerenders.
But as general practice, you’ll get the best results hey editing in a sequence that matches your highest res footage or output needs. If you want to have a 1080p output then edit in 1080p. 1440×1080 isn’t true 1080p. It’s an acquisition format that is trading off anamorphic squeeze for a better data rate. It’s going to be unsqueezed at some point for viewing, so why subject all the other elements of your edit to that.
I would assume that if you edit at 1440×1080 then that is the pixel dimensions that are used to upscale to 1902x 1080 on export. So even if your graphics were created at 1920, they’d be squished into 1440 and then that would be up scaled on output to 1920.
The differences might be slight, and it might be worth a more extreme test. Copy and paste your timeline into a 720×480 anamorphic 16:9 timeline. Export that as 1920×1080. See if it doesn’t look like SD upscaled to HD.