There seems to be a problem with classification images. The DDV image shows a lot of pixels classified as DDV, but the legend in SNAP says 0.000% DDV-pixels. It is unbelievable.
The same occurs for SCL-image. The numbers provided for different classed do not agree with the numbers provided in L2A-metadata. I checked and my computation agrees with the numbers given in metadata. There seems to be anything wrong with the total number of pixels used as reference to compute the percentages.