> If the vaccine was 90% efficient, you would expect the vaxed death to be 40%, not 84% (84% implies a 15% vaccine efficiency).

I'm not too sure how you arrived at your figures, but here's my back of the envelope calculation for the percentage of either expected deaths or expected cases coming from the vaccinated part of an age group:

  V * (1 - E) / ((1 - V) + (V * (1 - E))

where V is the percentage of the age group that's vaccinated (~95% for >70yo) and E is the effectiveness against either death or symptomatic disease (which isn't exactly the same as a case, I know), i.e. only (1 - E) of the vaccinated population will face death or symptomatic disease.

Playing around with the effectiveness values, 87% of cases and 84% of deaths from the vaccinated in the >70yo group doesn't look impossible.