computer analysis glitch!?

opponent's ...g6 leads to forced mate but was not considered a blunder. maybe horizon effect?
http://en.lichess.org/j5oLqE5e/

Something might be up with tagging of inaccuracies, mistakes, and blunders. 0/0/0 shouldn't add up to 46 centipawn loss and
here, http://tr.lichess.org/ASTOx7JS/white 4/2/1
shall make much more than 18 cpl. Both are recent analyses i believe

Sure, there is some bug with analysis.

It reanalyzes correctly here: http://en.lichess.org/snwBMnuX
Probably because it just made on other server