It seems obvious the chart starts in September so the reader can compare like to like. If the chart had started in January but ended with the most recent availa

It seems obvious the chart starts in September so the reader can compare like to like. If the chart had started in January but ended with the most recent available month (September) that would dishonestly bias an endpoint comparison since January is almost always colder than September. So the article text might be misleading - not a rare event with newspaper articles on technical subjects - but the chart is fine.