I've done multi-button with long press (and timed no-button auto-backout) on a '88 using ADC to save pins, and it worked fine, but field EMF noise can wreak absolute havoc on your results (fluorescent lights much?) if you're not using twisted, and preferably shielded wire. That could be what's throwing you out of your "consistent" results needed to qualify for a long-press. Ceradiodes or double-zeners can bleed off freak transients, and a capacitor at the chip A/D input pin would help smooth things out.