I think the point is that it LOOKS as if a child is learning to read by sight - and to start with, in books with limited vocabulary, that may be what they are doing , memorising words as wholes.
However, to properly read - attack an unknown word with confidence- anyone HAS to break it down. Whether that ‘breaking down’ is into smaller words, syllables, graphemes, that breaking down has to happen, and has to match with the reader’s ‘stored set of knowledge about what those bits sound like’. So faced with a word like ‘quadratic’, a reader may break it down ‘quad’ ‘rat’ ‘ic’ or ‘qu’ ‘a’ ‘d’ ‘r’ ‘a’ ‘t’ ‘i’ ‘c’ - and sound those bits out.
’Look and say’ requires a beginner reader to infer the breaking down process and what the ‘bits’ sound like. Synthetic phonics teaches that process explicitly - and is therefore successful for more children IF DONE WELL.
It’s worth pointing out that no process will yield anything more than a ‘best estimate’ unless the word is in the reader’s known vocabulary (ie they have heard it, or a word sufficiently like it to home their ‘best estimate’). I still remember the agony of being corrected by an infant teacher for reading ‘automobile’ as ‘auto’’mobile’ (as in above a baby’s cot).
This is one of the reasons why for sone children entering Reception now, the phonics-for-reading process cones too early. Some do not speak well, and have never been talked to. Many have very limited exposure to vocabulary. A lot are not able to actively listen precisely. The Phase 1 of the old Letter and Sounds is ignored at our peril - it focused entirely on listening and sound discrimination, without which teaching phonemes snd graphemes is premature.