Hi,
that's easy to explain. The datastore/blobstore deduplicates large binaries. So if you upload a large binary twice, it's stored only once in the blobstore. But the diskusage report fills it twice, thus resulting in a higher number than the actual disk usage.
Jörg