You're absolutely right. This slashdot article is a bad summary of a bad summary. The 8% number is in the abstract of the paper, and they cite human genome paper (from 2001) as one of the sources for that. What they're talking about is probably the overall % retroelement content, and not just the particular class of virus that this group is talking about. This is misleading because they also say the virus they're working with infects neuronal cells, which are more or less terminally differentiated, and so sequence from that shouldn't really end up in the germline cells. Of course, I haven't read beyond the first few sentences of the abstract, which is about the same that I think the PR person who wrote the article, and the submitter of the story did as well.