The data are the black dots and the optimum model is the gray line. Obviously, one would like to see all of the points on the line. However, the tails of a dataset hardly ever fall on this line. This is to be expected since, by themselves, they constitute a very small sample and small samples have large variance.
The plot shown here is, in fact, quite acceptable with the K-S statistic in the 69th (bootstrapped) percentile.