The decision was made to map these characters to their full width equivilant. I am aware that the characters in this set are pronounced differently to their full width counterparts but until a better transliteration method for the Ainu language is found, I figure this is better than dropping the charset entirely.

Interestingly enough, the only two nodes in the entire Japan dataset that use these characters are not even using it correctly (in my opinion). See:

From To Name Count
ㇰ ku KATAKANA LETTER SMALL KU -
ㇱ shi KATAKANA LETTER SMALL SI 1
ㇲ su KATAKANA LETTER SMALL SU -
ㇳ to KATAKANA LETTER SMALL TO -
ㇴ nu KATAKANA LETTER SMALL NU -
ㇵ ha KATAKANA LETTER SMALL HA -
ㇶ hi KATAKANA LETTER SMALL HI -
ㇷ fu KATAKANA LETTER SMALL HU -
ㇸ he KATAKANA LETTER SMALL HE -
ㇹ ho KATAKANA LETTER SMALL HO -
ㇺ mu KATAKANA LETTER SMALL MU -
ㇻ ra KATAKANA LETTER SMALL RA -
ㇼ ri KATAKANA LETTER SMALL RI 2
ㇽ ru KATAKANA LETTER SMALL RU 1
ㇾ re KATAKANA LETTER SMALL RE 1
ㇿ ro KATAKANA LETTER SMALL RO -