Science project - Electricity generation, temperature and sunlight hours in winter and summer

With a couple of days of summer holiday left, my oldest son, aged nine, began his project, or jiyu kenkyu--literally free research although if it were really free, then I'm sure he wouldn't have done it. His mum had been encouraging him to do something on flags of different countries, but in the end he decided to look at solar generation and the weather. This was partly inspired by his cousin, who was going to look at the amount of electricity and the amount of water they use. It wasn't my idea, although I was very happy to help him. It seems that these projects are more for the parents than the kids, and it felt like I was putting a lot more time and effort into this than I have done for my own research projects that have gone on to academic publications.
His first idea was that solar power generation depended on temperature, so he set about making a graph. One set of points was the the daily generation, which I've been copying into a googledoc spreadsheet from the display in the house, with his occasional help. Another data set was the temperature, courtesy of tenki.jp, who show nice monthly calendars with daily highs, lows and weather, with suns for sunshine, clouds for cloudy and so on. Because he wanted snow, he chose February first, then he decided to compare summer and winter, so he did another one for July later. He'd added the weather onto the February graph, which meant four lines running along the month, making it almost unintelligible with too much information.  The weather was very blunt, simply showing cloudy, sunny or snowy as of 3 pm. What we really wanted was the hours of sunshine.
It took a while to find data for sunlight hours online. Tenki.jp has the data but won't show it. Their live amedasu page has the last 12 hours for Matsumoto, but when you click for past data, it switches to Nagano city, which has different weather, on the other side of a mountain range, lower in altitude and closer to the influence of the Sea of Japan. We're in the mountains so even the next village has a different weather system.
JMA, Japan Meteorological Association, jma.go.jp, has extensive downloadable data, all the way back to 1872. Here they have nice monthly charts with the vital statistics for each day.
This allowed another two graphs of power generation (below in green) against hours of sunlight (in orange), which immediately appeared to be following each other, perhaps unsurprisingly.
His final report also showed graphs of power generation against temperature for July and February (daily highs in red, daily lows in blue). The power on the July graph seemed to be following the high temperatures, while the February graph power seemed inversely proportional to the low temperatures.

This makes sense as fine summer days are usually sunny, and cold winter nights are usually clear and followed by sunny days.
Next there was a table of the highest five days generating in February and the highest five days in July. The five best days in February were all over 50 kW hours, while the best five days in July were all under 50 kW hours, although the hours of sunlight were comparable. The averages were 52.1 kWh, 9.8 hours sunlight and 5.5 degrees maximum temperature in February, and 49.0 kWh, 10.2 hours sunlight and 32.5 degrees in July.
So it looks like, rather than higher temperatures meaning more generation, given the same hours of sunlight there will be more generation in Winter.
This is consistent with solar panels being more efficient at lower temperatures, which I think is because the resistance goes up with temperature. This was beyond the scope of this fourth-year elementary-school summer project, as we'd filled up the two pages which had already been a massive effort for all of us.