Since I'm:
a ) anal
b ) worried that I'm talking rubbish
c ) recklessly indifferent of being charged with egregious Pixel Peeping
I prepared the following crops using different sharpening.
Each was prepared from a fresh conversion of the original RAW file using ACR 4.1. All settings remained the same throughout except for sharpening. For the PK-sharpened versions the sharpening settings in ACR were set to the far left (Amount:0, Radius:0.5, Detail:0, Masking:0), which should mean that ACR is doing no sharpening at all (or at least I should jolly well hope so). I flattened the image directly after applying the PK Sharpener without altering the default fade amounts (which may not be optimal). The ACR-sharpened versions differed only in the sharpening settings specified.
The aim here is to get the bars in front (which are in-focus) sharp, while the out-of-focus leaves behind remain smooth.
Photokit Capture Sharpener, Digital High-Res Sharpen & Smooth, Medium Edge Sharpen:
Photokit Capture Sharpener Expert, Digital High-Res Sharpen & Smooth, Medium Edge Sharpen. The masks for both light and dark contours were modified by manually adjusting the levels so that the black level was set to 50 (In reality I'd have altered the gamma a bit as well, so this isn't really fair. But for experimental purposes I wanted to alter only one variable at a time.)
ACR 4.1 Sharpen, Amount 50, Radius 0.9, Detail 0, Masking 0
ACR 4.1 Sharpen, Amount 50, Radius 0.9, Detail 0, Masking 100
ACR 4.1 Sharpen, Amount 50, Radius 0.9, Detail 100, Masking 0
ACR 4.1 Sharpen, Amount 50, Radius 0.9, Detail 31, Masking 72
Blow them up and take a look, then post back about how I'm an idiot for worrying about such minor differences (that would set my mind at ease, if not my ego). After I'd saved everything as TIFFs I realised that imageshack autoconverts TIFFS to PNGs, but I hope that doesn't screw things up.
What does this prove? Well it proves that the Canon 28-135 IS lens I used to take the shot has a fair amount of chromatic abberation. It needed a lot of correction in ACR, but DXO cleaned it up almost perfectly, though don't get me started on DXO's 'Lens Blur' sharpening, just turn it off. (These are straight ACR RAW conversions, since I wanted to introduce as few variables as possible.)
Personally, I'd have to say that the final setting is the one I'd go for as a Capture Sharpening pass, though this is biased by other elements in the image that have been cropped out.