Strange asmfish behaviour?

[giovanni]

[d]8/8/8/4p3/4Pp2/4nPpB/6P1/5k1K w - - 116 71

I am puzzled by the behaviour of asmfish in this position. It thinks that 1)Bg4 leads to a draw, but realizes one ply later that it is instead mate in 6. Also, if you show it the winning variation 1)...Cg4 followed by 2)fg4 f3 and then you go back to the original position, it still thinks that Bg4 is a draw, i,e, there is no hash table learning.

[F. Bluemers]

check the movecounter,its a draw.
note that if it was not a draw by the 50 mvs rule,your suggestion Bg4 would get white mated after Nd1 and Nf2
Best

[giovanni]

My bad. Thanks for pointing out my faulty reasoning.