It occurs to me, after a leisurely breakfast, that the only reasonable way to figure the vehicular AM cores are to assume some sort of passive containment system.
What we're talking about is some form of Unobtanium© which creates a magnetic matrix, probably in a crystaline matrix. During its formation, its injected / innoculated with AM hydrogen, each atom suspended within a crystaline / magnetic cell, and requiring no power to maintain seperation. Liberating power from the system probably involves physically damaging the matrix, liberating the AM in a way that propels it away from the remaining coherent fields and toward what they hit the matrix with, which then reacts with the AM and liberates a relatively small quantum of energy (comparitive to, say, a full-bore unconstrained AM detonation; the energy needed to liberate the AM has to be a tiny fraction of the released energy to be worth the effort).
Juggle the numbers on the volume of the matrix material vs the volume of the cell vs the volume of AM contained to get numbers which fit the size of the AM cores on the templates. They specify "micrograms of AM," which is reasonable.
Of course, they have to be using "combustion" in some sense which we don't in regards to their "ICE Engines" (shades of BattleTech), since on Earth "combustion" is about materials combining with oxygen energetically. Since AM really doesn't require any specific form of reactive matter, oxygen or nae, they can't be using it in the conventional sense. If we assume they use the term to mean any energetic release from a physical matrix, it makes some vague kind of sense.
As dan / cal suggests, it probably makes more sense from a hard SF perspective if the fuel cells and patently not antimatter, because of all the points he makes. If its important they be so, we can certainly construct reasonable things around that, but its harder than it looks, in part because of observed failure-modes in the game as we clearly observe.