Solver settings in Abaqus/S for delamination analysis with cohesive elements

Hi 
 
I have different 3D models of laminates where I analyse the delamination propagation using zero thickness cohesive elements in Abaqus/Standard. My experience with the RIKS solver and Newton solvers are that they are extremely sensitive to even small changes in the step size settings as well as other solver settings. 

I have an example where a 3D DCB model cannot be solved using RIKS with an initial increment size of 0.01 and adaptive step size. But if I change the initial increment size from 0.01 to 0.011 it solves the model. The crack growth is first initiated at approximately step 0.05.

 
I think that maybe the solver in abaqus is not that suited for these type of problems or maybe it is the default settings that are not good for this type of analysis. I have read papers where the sub-plane approach proposed by Alfano and Crisfeld (2003) "Solution strategies for the delamination analysis based on a combination of local-control arc-length and line searches" has ben used with succes. An example is OVERGAARD, LUND and  CAMANHO 2010 - A methodology for the structural analysis of composite wind turbine blades under geometric and material induced instabilities. It seems that the solver is more robust and is able to solve for highly oscillating reponse curves. 

Does anyone know a good description of getting more out of the solver when solving delamination with cohesive elements?

Thanks in advance
Regards Brian Bak