DxOMark score does not depend on the optical scores.
It depends on the measurement perfomed for Sharpness, Distorsion, TStop, and shading. Noise, dynamic range and Color depth are taken into account too.
But in this case both lenses were tested on the same camera (ie, noise and dynamic range are the same). Differences could come from the fact that the separate scores (DxOMark Score, sharpness, transmission, etc.) are not calculated at the same f-stop for a given lens. For example, if the vignetting score is based on the maximum vignetting (ie, wide open) whereas the DxO Score is valid for the best f-stop of a lens.
But looking at the data in detail that doesn't seem to be the case here, the Sigma is better at basically all f-stops. What might be the case here is differences in how things are averaged over the whole frame for the individual scores and the DxOMarks Score. The Sigma drops resolution earlier and deeper when going towards the corners.