To do demand/supply matching, I don't use cumulative load-duration curves, because they lack the time dimension.
Given the sensitivity to Concentrating Solar Power (CSP) of time-of-day, what I'd look for is hourly or more frequent (eg half-hourly) data both for Direct Normal Insolation (DNI), and for electricity consumption.
Some utilities do publish such (half-) hourly electricity consumption; as do some transmission operators.
As for the insolation data, often reanalysis data is 3-hourly or 6-hourly, which might be too infrequent for your purpose. Reanalysis data does have better geographic coverage though. If need be, you might have to use more frequent met-station DNI data, and use that to interpolate the reanalysis data for your putative CSP sites, to get the right combination of precision for both time and place.
Combining data across timezones
When combining data across different timezones, the first thing (after basic cleaning) to do is to convert the timestamps in all the data (DNI data, demand data, everything with a timestamp) so that it's all for the same timezone (typically UTC, aka UTC+00). This requires adjustments for daylight saving time too. Once all the data are in UTC with no daylight saving time, then start doing your accumulation. (If everything's in the same timezone, then it's simplest to keep it in that ... except dealing with daylight saving time can create problems. As soon as I'm doing things across timezones, I switch to UTC as that keeps it all simple, and is most straightforward for any present or future collaborators, wherever they are in the world. )
further issues with synoptic- (continental-) scale analysis
One usual method with such synoptic-scale analyses is to first model without any transmission capacity constraints; and then to introduce such constraints later if it looks like they might become relative. Long-distance transmission is pretty cheap and easy compared to building generation or storage, so ignoring capacity constraints is a pretty reasonable assumption for a first-order estimate.