1920×1080, the standard HD 16:9 resolution, is about 2.07 megapixels. The first set of specs lists the sensor as having 2.29 megapixels for both still and video in that aspect ratio.
The later specs give you the actual files output. The "L" resolution is 4032×2272, which is 2.1× of 1920×1080 in each direction. And, hey, half of in each direction gives 2016×1136 — which matches the 2.29 MP from the sensor spec.
One of two things is going on. Either:
- The output image size is scaled up by 2×, which will not give any increase in image quality and in fact just wastes space, but makes the spec sound more impressive,
- Or, the sensor is actually 4032×2272 in raw sensels, but since these have color filters and it takes a grid of four in the standard Bayer layout to get color information for a single pixel, they counting that block as an "effective pixel".
The former seems more likely, and articles like this one confirm that this is standard practice in camcorder specs. So, you'll probably get better results by using the ""S: 2.1 megapixels 16:9 (1920 x 1080)" setting — although I'd experiment to see if this is cropped or scaled down from the apparently-native 2016×1136. In fact, because camcorders are not really optimized for high-resolution stills anyway, even if the second is true, you may find you don't gain anything in terms of real detail by using the higher resolution.