The flux will be stronger in the portion of the shoe under a stator pole piece than it will be under a gap between them. That means it wiggles around through the rotor pole piece's surface as the rotor moves from one stator pole to another. Field dragged through a conductor generates eddy currents, and those are the "surface iron losses" Flux referred to.
Flux: Won't you get the same effect in the magnets themselves, which are also conductive? Isn't the "surface iron loss" a symptom of the reduced gap (making the "field wiggle" take place more within the conductor rather than within the gap)?
Of course the losses increase with increasing conductivity. So if iron is more conductive than the neo alloy (I'm not sure if it is) the losses will be greater if the narrow gap is achieved with a curved magnet than whit pole shoes.
In any case: once your magnets are strong enough that you're saturating the stator you don't get any gain by reducing the gap further. So doing that at the cost of increasing the surface iron losses is a net loss.
I'd use shoes (and damn the surfaces losses) with weak magnets. But with big neos I'd assume I'm getting to stator saturation or near it, and say "Why bother?" Or try it both ways and see if I get more power out at a given RPM with the shoes (and then need to make a call on whether the drag and/or its heating's limit on furling point is more of a loss than the power gained).
In any case: If you use magnets with screw holes you'll want to use a non-magnetic stainless steel screw to hold the magnet down, to avoid "shorting out" part of the field.