Why Does a Part That Indicates True in the Vise Still Cut Out of Square?

Because "true in the vise" only tells you the part is square to the vise jaws, not square to the machine's own axes. If the vise itself is skewed relative to X and Y, or the part shifted after you indicated it but before the first cut, the part comes off the table out of square no matter how clean your indicator reading was.

The vise isn't the machine

Indicating a part square to the vise proves the part and the jaw agree with each other. It proves nothing about the jaw's relationship to the table's travel. If the vise was bolted down slightly rotated, or a chip got trapped under a corner during setup, every part squared up in that vise inherits the same offset. It'll still indicate clean every time, because you're measuring against the same skewed reference.

Check this by sweeping a solid ground edge or a dial test indicator down a fixed jaw face while running the axis, not just indicating the part in the jaws. If the readout drifts as you traverse, the vise is the problem.

Clamping can move what you just indicated

A part that reads dead true before clamping can walk when the jaws close, especially on thinner stock or anything with a rocking point near a parallel. If you indicate, then torque the vise handle down hard, then start cutting without a final check, you're trusting a measurement that predates the actual clamping force.

Re-check after clamping, not just before.

Work offset math can hide a real skew

If you're using a probe or edge finder to set a work offset from two points on the part rather than referencing the vise itself, a rotational error in point selection will bake a rotation into the offset. Two points too close together, or points on a feature that isn't actually parallel to the axis you think it is, both do this quietly. The G-code runs exactly as programmed; the offset was wrong from the start.

What to actually check

Work through it in order. Sweep the vise's fixed jaw against machine travel first. Re-indicate the part after final clamping torque. If you're setting rotation from probed points, use points as far apart as the part allows, to minimize angular error from any single bad point.

A vise that's true to the table, and a part that stays put after clamping, is what makes "square in the vise" mean anything. DigiForge's tolerance callouts follow ISO 2768-m as standard, down to ±0.01 mm, so it's worth knowing exactly what the print is asking for before fighting a setup to get there.