1) The 24:1 figure is just the inverse of the ratio. For example, if the output size is 1/3 the size of the input, 0.3333, then the inverse is 3:1, which corresponds to the idea that every 3 bits of input were (on average) compressed to 1 bit of output.
2) Every non-empty input file can be compressed by use of a suitable algorithm -- an algorithm that might only have gains for that one particular file. The question then becomes not whether .jpg files can be used as compression input (they can): the question becomes whether you can find an algorithm that compresses the subset of .jpg files that are interesting and useful for your particular purpose.
