The problem I had is, in the kanji transform mode, it seems to be unable to recognize many input, such as the half-sized characters (ぁ, ぃ, ぅ, ぇ, ぉ, ッ) and the voice with two characters (しゃ, しゅ, しょ, etc). In addition, the input method seems to only transform per kanji only, so words with more than one kanji need to be typed one by one. BTW, is there Anthy for MSCIM ?