this relates to problems with old romanizations, they took the examples of the syllables "na ni nu ne no, ka ki ku ke ko, ma mi mu me mo", so they simply assumed that it was the same for all syllables starting with consonants, hence: "sa si su se so, ta ti tu te to" etc., which turned out to be false, as there are no syllables that sound like "si, tu, ti" in Japanese, modern romanizations corrected the mistake so now those syllabels are more accurately romanizaed as "sa shi su se so, ta chi tsu te to, ha hi fu he ho" among other corrections.
It's not really related to the oldness of the system in question, Hepburn (the one with "shi") is actually the oldest standardised scheme. It has more to do with multiple systems being created with different design ideas (mostly about what is a "consistent" mapping from Japanese to Latin alphabet). But that doesn't change the fact that 1) all systems other than revised Hepburn are retarded 2) Japanese are almost universally taught kunrei/nihon-shiki in schools, so they tend to go with retarded options when romanising.
It has to do with whether you want to be phonemically correct, or phonetically correct. In Japanese [ʃ] ("sh") and are allophones, both mapping to the phoneme /s/. In other words, if there is an /s/ before an /i/, it always turns into a [ʃ], otherwise it stays an , but in the Japanese mind [ʃ] and are equivelant. You see the same thing with [ʧ] ("ch") -> /t/ and [f] -> /h/ ("hu" vs. "fu"). So when you do things like making し romanize to 'si', you are being phonemically correct.Because the phonemic form of the word isn't what you hear though, it usually looks retarded. Hepburn is more intuitive, and more accurate to the phonetic form of the language. It actually spells out what you hear, rather than the underlying phonemic form. I'd have to agree with 葉月 that it also makes more sense in my mind.
