At which point does increasing the cost of the converter reduce overall economy? That is, will the 5W saved by pushing to 97% be worth the 500 cm^2 of panel area saved?
Are there other methods to improve efficiency or economy (better panel cooling? better optimized mechanical mounts, wiring, routing, fusing, busing?), and what is their relationship in the same tradeoff?
Likewise for NRE, is the design time worth it, amortized over product life?
Come to think of it, at current prices, going to 97 or 98% is probably quite justified. 99% or more, you'll be hard pressed to approach, expected design cost going up hyperbolically...
As for techniques, you may find it's worth using a lower frequency, to keep switching and core losses down. You can afford to use rather large capacitors and inductors here. Quasi-resonant commutation may be helpful, but don't expect good yields on switching energy (like, maybe you can reduce switching loss by 50%, 80% even, but eventually you will be limited by the high frequency losses of the snubber components, which have to operate at high rates of change).
Tim