It sounds like your colleague does not understand the difference between estimate and commitment or it hasn't been communicated to him during training. And, since you tried to attach the problem to his personality, it's possible that your whole team doesn't yet understand it. (But don't worry! Most of our industry doesn't understand it. Agile is hard!)
When we say a story's size is X points, we actually mean a probability distribution. If our estimates are correct, the story should take longer 50% of the time (and the other 50% it will take less time). If your colleague believes that, when X units of time have elapsed, he will be asked to demo the story or else, that changes his approach to estimation.
Planning poker introduces another error: instead of trying to pin down X, we match it to a discrete scale, the Fibonacci scale (1, 2, 3, 5, 8, etc.) being the most popular. It is saying what the size isn't as much as what it is. When we say the story size is 3 points, we really say "it's X plus-minus some variance and X is closer to 3 than it is to 2 or 5."
Your team could benefit from understanding how imprecise this exercise is and how estimate differs from commitment. If you want/need to study these concepts in depth, this book has that.