While Tranpose() permutes the indices for dense UniTensor, such that the matrix representation is actually the transpose of the original matrix representation, this is not the case for symmetric UniTensors (or possibly all UniTensors with directed bonds). For the latter, only the direction of the bonds is inverted, but no permutation is done.
The permutation should be consistent.

The same happens when using Dagger().
Btw, the name .Dagger for conjugate transpose is very natural for us physicists, but I have not seen it in programming languages/libraries. I know .Ctranspose() or .H() (numpy). Maybe we should stick to a more common convention like numpy.