I’ve been using a standard script to build an alpha helix from an
alanine chain. Modeller generates the restrained structure but it isn’t
alpha-helical when viewed in PYMOL.
The restraints are consistent with an alpha-helical structure but they
are also consistent with other structures (they are quite loose). They
are designed really just to "encourage" a structure to adopt an
alpha-helical-like conformation but will typically not force a perfect
alpha helix.