It is notable that a number of teachers have reported their surprise that some of their better readers did not read the screening check's non-words accurately.
A number of these teachers suggested that the children were not able to read them because they were trying to make the non-words into real words.
I have a different hypothesis.
I suggest that this was not neccesarily the case. The children were told in advance that the words were not real words - and to aid this understanding, little drawings of aliens were shown alongside only the non-words. There was no reason why they should have tried to make the non-words into 'real' words in these circumstances.
Children who tend to 'read' books from a variety of cues such as picture cues and context deduction - and who may strive for fluency and perhaps apply their own level of common sense - may not be attentive to the details of words from left to right and may not use a blending strategy routinely or at all.
So, when faced with any form of reading, they may instinctively take a good guess. It is less likely with previously unseen non-words that a quick guess will be accurate enough - hence apparently good readers getting the non-words wrong.
The very same children may be perfectly capable of reading the non-words accurately by slowing down, applying their blending skill from left to right - in other words, increasing their attentiveness and applying the phonics they have been taught.
I have heard this is indeed the case when such children were asked to read such words after the event.
Children who read books should be just as capable of reading words in lists - they should not be context-dependent to read the words.
Children with a 'multi-cueing' type of reading profile are simply not used to blending all-through words routinely and automatically - their default strategy is more likely to be the guessing of unknown words.