It's just an example of lousy reporting. But first, some background. Whether a 'particle' or 'wave' approximation is more accurate depends on energy density. When the density of energy is relatively low compared to the energy of the photons (such as gamma rays coming off from a sample of radioactive material), the particle approximation is far more appropriate. When the energy density is high relative to the energy of the photons (like in a microwave oven), the wave approximation fits better. This is what was traditionally meant by 'wave-particle duality'.

This idea is indeed connected with the fact that the more localized a wavefunction is, the more spread is spectrum of momenta is (uncertainty principle). This is widely known and is nothing new. What they've done in this paper - which despite the lazy reporting of the paper is actually a thought-provoking bit of work - is consider 'wavefunction collapse', which is just the process of entanglement of the 'observer' wavefunction with the 'experiment' wavefunction. They've essentially shown that the amount of information that can flow from the 'experiment' to the 'observer' when the wavefunctions become entangled has an entropic bound. This idea has been thrown about for years; here they claim to have finally found a satisfying proof.