Actually, the data looks fine. The problem with that graph is that the y-axis scale makes it hard to see the actual change. Eyeballing, the temperature change over that time series appears to be close to a full degree of increased temperature, which is consistent with what has happened with the global average over the given period of time.
It's poor practice to create a graphic like that. Most of the range in the y-axis is unused. A better graph would have ranged between +/- 2°C for the anomalies. Since they don't actually post the actual trend, the coarse range makes it more difficult to see what is actually happening.
Now at American Thinker, they submit that choosing a sensible scale like that would be alarmist. But in reality, if the actual trend was given, then it doesn't actually matter what the scale is. The trend value wouldn't change, obviously. It's the data that determines the trend value not the scale of the axis. Further, their justification for choosing the +/- 8°C range is pure bunkum. They chose two sites on the globe, where the variability is high. But the data in the graph is for global temperature, which has has far less variability.
Par for the course unfortunately. We deserve better skeptics.