millermachine --
Don't forget to consider mirror thickness when selecting an autocollimation mirror, in addition to its planarity and metalization. For most "machine shop" applications of an autocollimator, I suggest a mirror of 2 to 3 inch diameter, at least 1/2 inch thick, flat within 1/4 wavelength of green light (flatter is better), and with a silicon monoxide overcoating on aluminum ("enhanced aluminum" is better) metalization. The type of "glass" isn't especially important, borosilicate is good, as is fused quartz.
Actually, the mirror doesn't need to be made of glass. Hilger & Watts mirrors are often steel, "diamond turned" aluminum mirrors are becoming somewhat common, and Starrett's Webber Gage Block Division makes superb autocollimation mirrors of chromium carbide.
Needless to say, it's vitally important that autocollimation mirrors not be strained by their mounting. If mounted mechanically, the mount-to-mirror interface should be as close to kinematic as practical, as should adhesive mountings. I'll point you toward an earlier thread discussing these issues:
http://www.practicalmachinist.com/vb/metrology/autocollimation-light-source-mount-211941/
In passing, I'll mention that the K+E magnet-backed parallel mirror, catalog numbers 71-6202, was described in the K+E catalog thusly: "A magnet back mirror is used to check the perpendicularity of any flat surface to a line of sight by means of auto-reflection or autocollimation. The mirror has a circular front surface 2 1/2 inch diameter. Its reflecting surface is flat to within 1/4 wave length of light over any area not greater than 50 mm diameter. A single magnet is cemented to the back of the mirror. Its contact surface is ground parallel to the mirror surface within 3 seconds of arc."
The K+E 71-6204 was similar but at 4 inch diameter was larger, and had three bonded-on magnets ground parallel to the mirror surface with 3 arcseconds.
John