The pressure is computed using the forces, so improved minimization leading to a lower potential energy (substantially negative) can directly affect the initial pressure.
By not minimizing using the same conditions, there can be a large potential change in going from a minimized model to the zeroth and first dynamics steps.
Assuming CRYSTAL is setup, you can easily compute the density:
! CONVERTS amu/A**3 to g/cc
set amu = 1.665
! ASSUMES, ORTH, TETR, or CUBIC; SEE crystl.doc FOR OTHER SHAPES
calc vol = ?XTLA * ?XTLB * ?XTLC
! PUTS TOTAL MASS INTO ?STOT
scalar mass stat
! COMPUTE DENSITY IN GM/CC
calc rho = @AMU * ?STOT / @VOL
It is possible to use the LATTICE keyword during minimization with the ABNR algorithm, but it may be better to adjust your initial unit cell edges manually, based on the pressure. Note that one can estimate the pressure after minimization, via something like
pressure inst temp 298. volu @VOL
assuming that VOL was computed as for the density.