I think the issue with any type of 'school for the gifted' is that academnic ability is a continuum, and isnot uniform across all subjects.
So children don't come labelled as 'gifted' or 'ordinary'. They come labelled as 'Exceptional at maths but just about middling in english ... and obsessively interested in football'.
Hence you have a problem of identification. What percentage is the 'right' one for needing different provision? What should you do about children with 'spiky' profiles who are either only very able at a single subject, or conversely very poor at a few areas while being very able in others? the 11+, for example, is notoriously poor at actually sifting out the genuinely top x%. A grammar which is supposed to cater for the top 5% probably actually selects a fairy random collection of the top 20%, favouring the best prepared and the most diligent and those from middle class households [and specific ethnic minorities].
A second issue is that even if you remove the top x% with absolute precision, the problem moves down to those in the x+1th percentile - the 'new outliers in the school for the non-gifted'. As there isn't clear water between the gifted' and 'non-gifted', being the 'now most relatively gifted' children now have the same problems that the top x% once had.
Tbh - as a gifted, year accelerated, prize-for-the-top-result at Oxbridge type - my perspective is that the issue is not being ABSOLUTELY gifted. It is being RELATIVELY gifted in a specific environment. So a moderately able child but in a low ability cohort (e.g. in a school in a particular catchment) will suffer more than a more extremely able child in a cohort with a full range of ability. So IME what would work best is balancing the admissions to all schools to make them genuinely comprehensive, so that all schools have much more closely matched cohorts to one another.