Unlucky,
I think that the point is that even if you 'appear' to be learning words by sight, many people who seem to learn this way actually work out the 'phonic code' and can then apply it.
My DS - a very early, predominantly self-taught pre-school reader - was fascinating to watch.
He memorised whole books by rote, and, being a phenomenally early riser, used to spend time reciting them to himself, following the words as he turned over the pages. He then learned to 'read' some words - as in, recognise and name them in other contexts - that he had seen in his familiar books, and this continued until he could read unknown texts fluently.
if asked at that point, I would have said that he earned to read through whole word recognition. However, in reception he joined the whole class phonics teaching - and his teacher found that he had an excellent knowledge of phonics for decoding, including e.g. digraphs and split digraphs.
So what LOOKED like whole word recognition was actually initial whole word recognition followed by working out the phonic code so that he could recognise the graphemes in other contexts and thus read unknown words.' Having seen it in him, i can now see that that is exactly what I do - I too was a scientist: I apply the knowledge of phonics from words that I already know to break down and then build up the new word from these segments.
Teaching phonics directly simply takes out the requirement for the learner to work out the phonics code for themselves - and thus makes it more likely that more learners will succeed. Why 'hide' the mechanism by never directly teaching it, and thus require every learner to work it out for themselves?