Maybe if we do that for the grammarian's re-wording instead, on both the -a of extincta and the other -a of deleta, then the thing fits into a ditrochee + a bacchius of sorts, after basically reducing the first -e- of deleta too (dēleō dēlētus):
A bacchius can be both u – – and – – u, so you don't need to shorten anything in deleta, nor do you need to lengthen the a. with a long a in extincta, it would simply end in – u – u / – – u
The idea of considering a nominative a long sounds rather weird though.
It's possible that by ditrochee, he has the scheme – x – u in mind with x being either –, u or u u ... in that case, you would have (ex)tincta sit atque deleta as – uu(=x) – u / – – u ... with the bacchius starting with a long syllable, as Laurentius said. If you give a ditrochee the – x / – u scheme, you would also get away with considering conservare a ditrochee.