I don't understand all the details but F5 stegnography encodes data by altering the DCT coefficients per 8x8 pixel block, those coefficients are stored with Huffman compression. The method of encoding is why the output image is always a JPEG. You would have to do statistical analysis of the JPEG coefficients… (assuming the software wasn't comprimised to leak additional info as well, the absence of JFIF header appears to be such a case)