As Ron said, it needs to be closer to the drive and the door. Detection range is 15 to 20 feet, for reliable operation, and best sensitivity is across the sensor. When the postman walks down the drive, it is across the sensor, good detection, but furthest away, probably 30 to 40 feet. When he turns to walk to the door, he begins to get nearer, but is moving directly towards the camera, which is least sensitive in that direction. He will also be closer to background temperature, whereas the lady coming out of the house will be very warm, moving into the cold, across the sensor, and as near as possible/likely, with your current setup.
If you are the same temperature as the background, the sensor has nothing to compare you to, and you may as well be camouflaged. Indeed in terms of your infrared signature you are camouflaged, because everything will be at the same temperature, apart from small patches, such as your hands and face, if you are nicely wrapped up, like the postman might be. In other words, the outer surface of his clothing, exposed to the atmosphere, will be at the same temperature as everything else, and only small patches will stand out to the detector.
Turn up the sensitivity, as Joel mentioned, and you might end up with every passing car triggering a response, especially at night with hot headlamps. Zone them out, and it may help, but again as Joel mentioned, it’s not a perfect solution, and even when working as it should, it still wastes battery, because the camera still becomes active for every movement, does a bunch of checking, discards the clip, then goes back to sleep. With a lot of traffic, and some wind in leaves, etc, it could do this hundreds of times per night, for example. Yes, better than hundreds of clips, but still gobbling up your battery.
Move nearer, and don’t have any window in view. What’s the point, of having any of your field of view looking at the side of the house itself? No one can be ‘in’ the wall, or the window, and if they are breaking them in, then they will be outside long enough to get a shot of them, plus you wouldn’t see them as they passed through anyway, so you may as well minimise the actual view of the property.
This would also help make the path a little more ‘across’ the sensor. However, it will also make the road more across the sensor, so it might be a ‘test it and see’ situation.
With all said by the three of us now, it may well be clear that waiting for real world events to cause a trigger, or not, and then moving it to see what happens, isn’t really practical. Perform active testing. Re-position, walk down your own drive, see what happens. Change the sensitivity, wait for a car, or a van, see what happens. Go to the door, etc, etc. In no time flat, couple of hours tops, at the weekend, and you will know all about what works and what doesn’t.