I noticed that you didn't use Helicon Focus' 'method B' (similar in segmentation approach to Zerene Stacker's Pmax mode). I have to assume that you chose the optimal parameters for the 'Method A' you apparently used, i.e. you didn't mention which settings for radius and smoothing gave the best results so perhaps you just used defaults?
Likewise, Zerene Stacker offers parameters to improve the results depending on the stack 'quality' (amount of DOF overlap).
Your results are interesting, in that an earlier test I did between Photoshop (CS4?) and Helicon Focus Pro turned out less favorable for Photoshop. Perhaps the different Photoshop versions give different results, and CS5 has improved.
And while I agree that it's desirable for the initial output to be close to ideal, it is also important if and how the (inevitable) artifacts can be retouched most successfully. Photoshop doesn't offer much in that respect.
You are correct that I used the Helicon default.
I don't know if there is a difference between CS4 and CS5, but the CS5 worked best for me by a pretty wide margin (especially in maintaining the quality of the bokeh). Regadring dealing with artifacts, I have found Photoshop's heal, clone, and lasso tools to be excellent in this respect.
If we call the composite stack the "Master" file, and if we call the individual files that were used to create it the "donor" files, I have found that using either the Heal/Clone/Lasso tools on the best donor file of that created that part of the Master image, and applying the tool to the area of the Master file that needs work, works wonderfully.