The screen resolution on your iPhone is 480x320 (960x640 on iPhone 4), so even a 2MP image is downscaled when displayed on the screen. Downscaling means that only one in X pixels is displayed, depending on the ratio between your screen resolution and the image resolution.
As you zoom in you gradually see more individual pixels but less of the overall image. Eventually you reach the 100% zoom where every pixel on the image is displayed by exactly one pixel on the screen. Zoom in past that and you will see that several pixels on the screen show a single pixel of the image - the image will look blocky. Your screen is now upscaling the image using the crudest form of interpolation - the zero-order hold - where every image pixel is displayed by a square of Y screen pixels. The program you speak of does the same thing, but it uses more sophisticated interpolation techniques so that the image looks better. One thing remains the same in both methods - you don't gain any new information and detail that wasn't already captured by the camera.
Now, take a 2MP image and an upscaled 5MP image, view them both on your iPhone screen at full screen and they will look exactly the same, because both are downscaled to the iPhone screen resolution. View them on a larger resolution screen, however, and you can benefit from the quality upscaling because it would not, ideally, be as blocky as the 2MP image. However, you should note that even a full HD display has, surprise, 2 million pixels, so it will be hard to find an actual display where you can use this benefit.
Note that all this time I was talking about a 2MP camera. The N900 has a 5MP camera already, so upscaling those photos is even more futile.
Hope I cleared it up for you.