Currently ZipArchiveOutputStream will add data descriptor record when the output do not provide random access. But if you add an entry using addRawArchiveEntry then the CRC, compressed size and uncompressed size could be know and there is no need for data descriptor record as those values could be set in the local file header. The current implementation does both - it sets the correct value in the local file header and adds additional data descriptor record. Here is the relevant code from ZipArchiveOutputStream#putArchiveEntry:
Maybe ZipArchiveOutputStream could be improved to not add data descriptor record when the CRC and size values are known in advance.