The short answer is: no.
When you compress your file into an archive (such as zip, 7z, rar, gzip,...), the algorithm is looking for ways to store the given information in a shorter way. This is for example done via Huffman coding, where the algorithm checks for the frequency of values and then stores the mot frequent value in the shortest number, the second most used in the second smallest number,... (this is extremely oversimplified - please read the linked wikipedia article for a better explanation. Also, of course, the Huffman coding is only one - very simple - example of compression). This also explains why compressing an archive often further reduces the file-size.
So your 75% file size reduction is because the compression tool finds a more efficient way to store the information than the video- and audioencoder did. However, usually, there is no way to tell an encoder how to store the information, as this is something that is specified in the standard. Sometimes a newer version of the encoding program can help because of better implementations of the standard, but usually, that just applies for brand-new codecs.
Your video-player, however, can't handle multiple compressions simultaneously. I think that this is mostly because it would need quite some ressources (both RAM and CPU) to decompress the archive-file and then read the file itself. Also, it would still need to decompress the whole thing first, just as you need to do, because it is quite difficult to estimate from a file you don't have where which stream is stored and so on and so forth.
So if you need to work with theora/vorbis, there isn't much you can do about it. Otherwise, if file-size is your priority, I would recommend looking into VP8 / VP9 / x264 / x265. If you want to go the VP9-way, I recommend something along FFmpeg's Encode/VP9-Guide (linked just above):
ffmpeg -i <input> -c:v libvpx-vp9 -crf 15 -b:v 0 -c:a copy <output>
<output> with the respective file-names (and file-extensions). Both *.mkv and *.webm work (I only tried these two), however, *.webm seems to require Vorbis as an audio codec (e.g.
-c:a libvorbis -q:a 6 instead of
-c:a copy), where *.mkv doesn't seem to have requirements.
-crf specifies a constant rate factor (imagine it as "constant visual quality"), where lower values mean better quality.
All of the above of course assumes that your files were encoded efficiently (e.g. not specifying a bitrate of 6 Gb/s) and with a properly working program. I for one have never been able to compress much out of any lossy compressed audio- or videofile (but I seldomly use theora or vorbis).