mpl 1.5.3/2.0b4/master, no matplotlibrc

_xs = np.linspace(0, 30, 201); plt.plot(_xs, scipy.stats.lognorm.pdf(_xs / 14, .472) / 14, ".")

Note the slight inaccuracy in positioning of the markers at the positioning of the markers at the positions marked with the red circles, resulting in a "step-like" behavior (you may need to open the image in a larger viewer to notice it).

The data itself does not present this behavior, as can be checked just by zooming in these regions.