Practice also applies to recall, and this would be closely related to context effects.If a person has heard a song 100 times without actually being able to decipher some of the words or understand their meaning, then that practice may be largely wasted. However, I think this should be qualified with the above points. Thus, the more times a person has heard a song, the better they should be at recalling. Practice is a fundamental effect in recall.There would probably also be a range of other effects related to properties of the lyrics which would make a given song easier or harder to recall (e.g., rhyming and rhythm in lyrics might make lyrics easier to discern). Recalling meaningful sentences that follow standard grammatical structure is much easier than recalling random words, and increasing the depth of processing (i.e., greater meaning) of the material should also help with recall (see this discussion of level of processing effect).

Even when the words of the lyrics are readily discernible, lyrics vary in the degree to which (a) they have easily discernible meaning (b) follow standard grammatical structures (c) repeat lyrics, etc.If you decipher alternate words, then this may reduce the degree to which the words have semantic meaning, and therefore the ease with which the lyrics can be retained and recalled. If you can't decipher the words, then you wont be able to recall them. Many songs have lyrics that are difficult to decipher.From an information processing perspective, I'd conjecture the following points:

Sainik movie song lyrics