Transitive reduction failure

[snowleopard]

The following event log triggers a bug in transitive reduction of the produced partial orders:

X a Y1 Y2 Y3 b c Z
X b c Y1 Y2 Y3 a Z
X Y1 Y2 Y3 b a c Z
X Y1 Y2 Y3 b c a Z

The following two partial orders are produced:

g1 = X -> Y1 + X -> a + X -> b + X -> c + Y1 -> Y2 + Y2 -> Y3 + Y3 -> Z + a -> Z + b -> c + c -> Z
g2 = X -> Y1 + X -> a + X -> b + Y1 -> Y2 + Y2 -> Y3 + Y3 -> Z + a -> Z + b -> c + c -> Z

Note, the only difference between g1 and g2 is the arc X -> c which is clearly transitive, as both X -> b and b -> c are also present.

The correct result should be just a single transitively reduced partial order g2.