There are two possibilities I can think of for getting different readings on different occasions.
One is that if a shim is worn, there might be an even depression, or else a groove matching the cam lobe, depending on how freely it rotates as the engine runs.
The other is if the cam was in a slightly different position for testing. We assume that if the valve is closed, the whole of the back rounded side of the lobe has the same profile - ie a perfect circle. That may not be true, hence the importance of using the rule about checking certain valves when others are fully open ("rule of nine", I think).