I can hear you ask why this should be a problem. It was because the link looked as follows:
In the end it worked out that document.evaluate became the saviour of the day, since it allowed me to make xpath queries on a document. The result is a java step definition and cucumber feature file that look as follows:
I suppose you could argue that the html could be changed to be better. True, but atimes that might not be an option. The purists will argue as they do here that it isn't a feature that should be provided by WebDriver but I think it would help to have a way of traversing the DOM once you have looked up an element. I should like to read about other creative solutions to a problem of this nature.