EDIT: A free pdf of a draft of the paper is at: http://arxiv.org/PS_cache/arxiv/pdf/1009/1009.6126v2.pdf. A quick skim suggests that at 14 qubits the state they actually prepare in the lab is, indeed, not very similar to the state they intend to prepare, with a reported fidelity of about 50%. That's the same fidelity they'd get if they just prepared an all |0> state. While the paper reports terrifically interesting work, this and many other details in the paper suggest quite a subtle picture.