Hi Adriaan:
You might find this link useful:
http://www.math.le.ac.uk/people/ag153/homepage/Gorlov2001.pdfIt's a paper by Gorban, Gorlov (yes, THAT Gorlov), and Silantyav, deriving limits (GGS model) similar to the Betz limit for substantially planar kenetic-energy harvesting turbines in unconstrained flows of INcompressible fluids (i.e. propeller-like turbines in water that's flowing freely, with negligible head, in a broad stream rather than a duct the size of the turbine).
They say Betz substantially overestimates efficiencies in water and come up with a limit of about 30.1% for essentially planar turbines (like propellors), though three-dimensional structures can go higher.
Your KD 598 describes a Garman turbine (propeller-like turbine on an angled shaft below a raft), which tend to run between 15 and 18%. You're a razor-hair under 15%, which is right in the ballpark, and good work for an easy to construct rotor (rather than something with a highly-optimized waterfoil shape).
The (then recently characterized) Gorlov turbine runs about 35% (and was the top of the line by a bunch).
It's essentially a water version of a helical darrieus, with supporting disks at the top and bottom that constrain the water from entering or exiting except through the blade-swept space. But for water the blades are much wider - about half as wide as the spaces between them. As a helical it doesn't vibrate substantially and self-starts. As a broad-bladed, high-solidity design it can harvest substantial energy from slow water flows.
Looks to me that a vertical-axis Gorlov should be easy to construct (using, for instance, the same technology as fiberglass or wooden boat hulls), would be sturdy, would hang straight down from, and be entirely under, a supporting raft, would harvest more than twice the energy as a propeller-like turbine for a given swept area and (being rectangular) would have a substantially greater swept area, and would probably move slowly enough, visibly enough, and with enough pressure disturbances to trigger the lateral-line sensors, to be detected and avoided by fish.
The main problem would be supporting it against the side-forces trying to bend the axle, which would require either an underwater bearing at the bottom of the shaft or a strong shaft and two separated bearings above water. (The Garman is easier to support from above because the rotor loads its shaft almost entirely in tension, so even if you do use a cutlass bearing underwater it only has a tiny load, while the Gorlov would need a bearing that could support half the drag.)
The paper doesn't give the details of constructing a Gorlov rotor. But it should be easy: Like a Darrieus, it should work well with any decent, low-resistance waterfoil shape that lets the flow attach to the front from a modest angle and attached flows leave the back tangent to the circumference and with negligible turbulence. With a vertical shaft the generator and transmission or belts would all be above water and easy to shield from both splash and rain.