Setting aside the concept of the core vs the surface, since I don't have an opinion on that. (It might be cool. It might not. I can't say.)
Most examples of magazine explosions (or, really, turret explosions) I can think of have severely damaged the ship in question, and many of them have fooled observers into thinking the ship was out of action (e.g. Seydlitz at the Battle of Dogger Bank). However, I can't think of any magazine explosions which have actually completely destroyed the ship in question.
With that in mind, I'd model it like the following:
- First, magazines designed with proper ejection technology have whatever percentage chance to avoid a magazine explosion at all when HTK on a magazine is successful.
- Assuming the magazine fails this role, it has a 10% (or other small, flat, non-negligible) chance for a catastrophic explosion
- Catastrophic explosions add together the warhead damage of all missiles remaining in the magazine in question and apply it to the internals of the ship. Bye-bye.
- Non-catastrophic explosions add together the warhead damage of all missiles remaining in the magazine in question, then multiply it by some fraction (e.g. 80%). This can be flat or tech-dependent, as makes sense. If tech-dependent, I would suggest it's the same tech as magazine ejection chance. This models that proper magazine design not only provides for emergency ejection, but attempts to channel as much of the explosion of possible outside the ship in the case of ejection failure. Then, the explosion is applied to the ship internals, as above.
To provide balance (and, I think, realism) with ships which rely on reactors, reactors should also have a safe-ejection technology which will mitigate the chances and effects of power-plant explosion, though a catastrophic failure could always happen, as with magazines above.
(And of course, 'always hit an empty magazine' is silly.)