The KANJI1 server character type is different because its collation is single level, based on the CollOrderCS and CollOrderUC fields. In other words, KANJI1 uses the second-level collation values as its only level. Because of this, it is not possible to define a collation that works well with both KANJI1 and other server character sets.
Note also that for KANJI1, only single-byte characters are affected by the MULTINATIONAL collation definition. Multibyte characters continue to collate as always by means of binary comparison of the internal representation of those characters. This should be very similar to how those characters would collate in the client character set they were entered from, with the exception that EUC cs3 characters on the client would sort before, rather than after, cs1 characters.