k-point sampling is often crucial for convergence, to get an accurate Fermi level. Same for the number of layers in the electrode, which often is chosen too small. Otherwise we have improved convergence a lot in 10.8, so if possible you really should try it.