Cardio machines are notoriously off (As much as 30%!) ; the only way to get an even close-to-accurate number is with a heart rate monitor (with a chest strap.) It's still just an estimate, of course, since the only way to be truly accurate is in a lab, but it's about the best we non-scientist types can get. ;)
When I have multiple numbers and I'm not sure which one is right, I either take the lower number, or split the difference. I'd rather underestimate than overestimate.