Been thinking about this one.
What % of children could be educated in a grammar-type school - so rather than thinking about having a 'life raft' for between 0.2 and 23% of children (depending in whether it is an area with residual grammars or a fully segregated system), how many children might be able to benefit if grammar schools could accommodate all who had the 'academic capability' to manage the curriculum there?
79% of Year 6 children attain Level 4s across the board - and in Kent at least, children who obtain Level 4s go on to grammar schools. Even in a county with superselectives only, I have personal knowledge of highly-coached children with L4s, or in one case not even a full set of L4s, passing the VR test for the grammar. As there doesn't seem to be a huge percentage of children 'managed out' of those ghrammars after 11, then children with L4s (at minimum, there may well be other children e.g. those with ESL who are very good at Maths despite less than L4 in English who could also thrive) are capable of managing a grammar school education.
So that leaves a maximum of 21% who, on this very rough estimate, might not be able to manage a GS education ... or maybe even fewer, once the inherent unreliability of any test at the margins, and late developers, are taken into account.
I am, marginally, in favour of extreme superselectives - in the same way as I believe that there are children with cognitive or other impairments who are best educated out of mainstream in special schools, I believe that there are a tiny % of children - certainly less than 1%, maybe as few as 0.01% - who are of such high ability that they cannot be educated in a mainstream secondary because their high ability is so extreme. Those children who need university level Maths at 12, for example, are not easy to educate in mainstream, and a statementing process such as those used to identify those needing special school placements could be used to identify them and place them in 'High ability special schools' at whatever age they are identified.
But if 79+% of children are at or above the minimum academic level at which grammar schools are currently successfully accessed, there seems to me to be very little argument for hiving off only some of this % on the basis of an arbitrary and unreliable (in the sense that it will not, on separate days, identify the same children as 'passing' or 'failing') test at 11..