I totally agree that this is human-computer-computer-human communication, and is far from direct. Not overly excited by any means. One thing that worries me is this comment from the Discussion:
This approach could lead to conscious synthetically mediated modulation of phenomena best detected subjectively by the subject, including emotions, pain and psychotic, depressive or obsessive-compulsive thoughts.
One thing that the researchers did right was send a short, understandable, message. Using a Bacon cipher, they encoded "hola" and "ciao" into the transmissions. Had the receivers been trained to read the code, they would have been able to understand the message - it was received correctly, and their receipt of code was recorded. Them not knowing the code is actually better proof that the transmission was valid. They all got the same transmission which (although it contained language) to them was gibberish.
Here's the direct link to their study, and if you read the Discussion and Supporting Info, you'll get more about that. It was glossed over in the article.