The article is pretty light on the details, so I'll try to add to it. In my prior career as a cognitive neuroscientist, I analyzed an fMRI data set on blindsight patient GY, and read up on the blindsight literature at the time.

The neurons in the retina don't go straight to the occipital (primary visual) cortex; they all synapse onto the thalamus first. From there, the vast majority of exiting neurons go on to occipital cortex, but a few percent go off to other regions, such as the superior colliculus, amygdala, MT and FEF.

It's been shown in diffusion studies that people with damage in thalamus or prior to it, don't show blindsight effects, while people with post-thalamic or occipital damage can.

Now, the really interesting part is that, these non-occipital signals are enough to help people perform tasks at above-chance levels, but not enough for them actually be aware. Nobody knows for sure why. Conscious awareness is one of the biggest unsolved mysteries of neuroscience. Top candidates are information integration theory, higher-order thought theories, and signal-detection-based theories.