It is absolutely rational to tell polite lies. If people want to hear the truth, they should first somehow show that they are the rare 10% who can take it as constructive advice, not get upset and angry like most people.
I have to agree with Jimmy Wales on this - this is experiment is just as "insightful" as demonstrating to people that you can get away with vandalism.
Yes, it's not that difficult to troll Wikipedia. Just as it's not that difficult to scam old people, dump your trash in the forest, or scratch cars in a parking lot. You would most likely get away with it, but it does not mean that there is a huge security risk in parking lots that the world needs to be made aware of.
Society is based on the fact that most of the time, most people are not assholes, and therefore we don't need a policeman following everyone at all times. People don't troll or vandalise because they see it as the wrong thing to do - and the small risk of getting caught, and humiliated or punished is sufficient to discourage the less ethical ones.
A 7% error rate is, for all intents and purposes, worse than no test at all. At least with no test, people aren't going to falsely believe that they're not infected. They can take precautions.
You make a huge, and in my opinion wrong assumption that most people believe "I could potentially be infected" or take precautions if they have not taken a test.
People are terrified of the thought that they may be infected, and for many the bliss of ignorance is more comfortable than testing. Many people think that there is no way they could be exposed, because it's something that only homosexuals and drug addicts get, and sleeping around with "nice random people from parties" is just fine. Some are even assholes who think that testing is a bad thing, because they suspect having something, but can always claim ignorance to new partners if they haven't been tested. Some are major assholes who will outright lie about this.
Perhaps these opinions are not universal, but they are the majority what I've heard expressed in college environments. A quick home test, even with a 7% error rate is much better than nothing: for people who might not take the effort of going to a clinic otherwise, and as given as a request to new partners.