We can think that the experiment wasn't the best one due to the low take-up. However, even taking this into account,  it is still thrilling that they didn't find the expected results as the concentration of treated workers increases by area. Maybe this could be driven because the quintuplets of similar areas weren't correctly designed. Hence, apart from the low take-up, distinctive characteristics by area could also explain this finding. 
Since it is a large-scale experiment in a western country, the external validity of the results can be considered to hold in countries with similar characteristics. However, it would be restricted to the targeted sample. Countries with lower concentration of skilled job seekers may have diferent results.