I have said before that I believe that grammar schools should essentially be treated as 'Special Schools' for those who have the special educational need of being so academically able that they cannot efficiently be educated in a comprehensive school.
In the same way that some children who have e.g. severe learning difficulties cannot be educated efficiently in a comprehensive because the specialist teaching, knowledge and equipment that they need are most efficiently used if they are concentrated into a single location to which pupils are transported, the same should be true of those who are academically so advanced that the teaching, knowledge and equipment they need (e.g. to access university level Maths in early Secondary, or a totally different curriculum than that for conventional 'GCSEs at the age of 16') can only efficiently be delivered in a few locations to which students are transported. Music schools like Chetham's, or dance schools like White Lodge are essentially a similar model, but based on 1 particular subject.
Ideally, these would also be co-located with normal comprehensives, so that if a child is advanced only in a single subject (Maths in particular is something that a child can be gifted at 'in isolation', while having very normal ability elsewhere) they could attend a normal comprehensive for the remainder.
The identification should be by the same process as for the transfer / entry of a child to an existing Special school - collaboration between parents and school to detail need and why the existing provision for them is insufficient, then a range of tests and pupil / school meetings administered by an Ed Psych.
Although existing Special Schools do have c. 3% of the school population in them, that is for a range of SEN, not just for learning needs. I suspect therefore that the 'Special Educational Need' of being so able in one or more subjects that a conventional comprehensive cannot meet their needs is much rarer - certainly less than 1%, possibly down at the 1 in 1000 level.
Yes, there are very highly selective grammars at the moment, who may take only 1% or so of the cohort, or even less where there is a very large effective catchment. However, they don't accurately identify - or need to identify - the actual top 1%. Instead they identify 10% of the children within the top 10%, as anyone in the top 10% can cope perfectly well with their curriculum because it is still predominantly based on conventional curricula leading to conventional exams - which is not what those who genuinely are able at an SEN level need out of their education.
I know one of those 'top 0.01% type' children - and their learning needs are totally different to the norm: as I said, University level maths in early secondary, delivered in their case by a mixture of a university tutor delivering Maths puzzles by post, and streaming of university Maths lectures. In other subjects they were simply top set in a comprehensive, which met their learning needs in those subjects perfectly.