Are you calling a M5? The way Mach now works (with the parallelport) is it needs to know the spindle has started, so you need to call an M3, likelwise it needs to know the spindle has stopped and thus you need a M5. Alternatively pressing the screen button for Spindle On/Off will do.
Also if you have averaging on in Ports and Pins, Spindle Setup, it may be a bit slower to respond to the speed. It will not affect threading as that is read directly in the driver, its just the DRO.
Hood