Iso is a container format; it’s a 1:1 mapping to how the bits are stored on disk with a header at the start describing the structure. Bin/cue separates the header into a separate file and can include data structured in ways that don’t comply with the ISO-9660 standard.

WinRAR is a compression/decompression program. It supports multiple archive types and compression formats.

Depending on the type of data you are compressing and whether you want lossless or lossy compression, you’ll want to select a different compression algorithm.

Depending on how you plan to use the files, you’ll want to use a different archive format.

Assuming you use the rar archive format, you still have a lot of options to consider. Should the data be encrypted or not? Should the directory structure be encrypted or not? Do you want parity files and segmented archives, so that if one of the parts gets corrupted (or goes missing), you can still extract the original data in a lossless way?

Beyond all that and selecting the compression algorithm that best compresses the type of structured data you’re storing, the general rule is that if you’ve got lots of data, using the largest dictionary and the largest compression window you can will result in the best compression.

So the dictionary is essentially a code book that says “when I see data x, represent it with data y in the file”. The compression window is how much of your original data is loaded into memory at any given time for the dictionary to look at and compress.

[edit] if you’ve got video, the best compression format commonly available today is H.265. This is a lossy compression format, meaning you’ll never be able to precisely recreate the original file. But it’s close enough not to matter at the right compression settings.

And if you’re using H.265, the best container format to stick it in is an MPEG-4 archive (typically with a .mp4 extension).

The result is a highly compressed and structured file. Loading chunks into memory for rar compression will usually result in a larger file, because the data is already compressed in a structured manner that a general compression algorithm can’t match… meaning that you’d get the input of xxxxx resulting in output of yyyyyy.