Super-resolution ultrasound imaging has evolved using image analysis algorithms. However, the images used are not generated with beamformers that are designed for single particle imaging, but rather for anatomy that provide continuous features (eg. delay and sum). In order to compare image- and signal- derived localisation accuracies we used multi-focal imaging combined with the simple metric of sharpness. A 7 MHz (λ=212 μm) linear array with 192 elements is used to scan a phantom that is composed of a thin wire. The average axial localisation accuracy using the sharpness method on the raw signal is ≈ 0.01λ while the centre of mass best measurement on image data provided ≈ 0.06λ. It is concluded that image derived localisation is compromised by the process that generates the image. It is therefore suggested that super-resolution imaging will benefit from alternative beamforming methods that are designed to enhance single particle imaging.