But if you are manipulating your curve when there is a genuine grade boundary already set as in this example surely you are doing your students a disservice to award them a higher grade
Not if you are talking about expected grades, not working at grades. If you are giving them a mock and want them to know what they would have got if it was the real exam, then using the actual grade boundaries is fine.
Predicting grades in my subject also involves ranking pupils, sometimes just mentally. If I'm teaching top set I know that top set gets grades between 7-9. So I'll give a kid whose test results put them at the bottom of the group a predicted grade of a 7, and the kid at the top a 9. It gets fuzzier in between.
Parents will then think I'm good at predicting grades because they're mostly right.
The problem we've got here is that we are not supposed to be awarding the kid what we think that they would have got in the real thing had covid not existed, which we did last year. We're supposed to be looking at their work and giving their work a grade. But grades aren't based on standards of work, they're based on rankings of pupils.
If we wanted to give students a reasonable 'working at' grade then we would give them a mock and use that year's grade boundaries. Except we can't do that because we're only supposed to be testing them on what they've been taught. So the paper will be cut down and the working at grades will go up, but in a way that won't be standardised across schools.
That's if you use past papers. If you're doing your own assessments, the grade boundaries will be either totally made up before the paper, in which case good luck, or they will be set once the results are in to ensure that the cohort get the 'right' grades (i.e. ranking pupils and fitting the results to the usual grade profile for a cohort in that school).