The spec sheet tells us that the 24K/W figure is using natural convection at 75 degrees C above ambient.
Looking next at the falling line on the graph, the arrows indicate that it graphs thermal resistance against air flow. We see that the line never reaches 24: the values of thermal resistance are much smaller for all the plotted air flow values.
Looking now at the rising line on the graph, the arrows indicate that it graphs temperature against power. Drawing a vertical line at 1 Watt, we read a temperature rise of 28C, indicating a thermal resistance of 28. This is much larger than any value on the Air Velocity curve: it is close to the quoted value for natural convection. It is not 24, because the line is a curve: 24 is only specified for one point on a curve, the 75 degree above ambient point.
Drawing a horizontal line at 75C, we read approximately 3.25W, indicating a thermal resistance around 21.
This indicates that (a) The graph is wrong: it should go through (75C, 3.125W), or (b) 24K/W was wrong, and the graph is correct, or (c) The power/temperature line is drawn for different conditions than the quoted 24K/W figure: different ambient, different mounting, different air flow etc.
In any case, as the graph shows, the value 24K/W is not exact for any other point, or any other conditions, but it is approximately true for condtions approximately the same as specified.