In the first example, it doesn't mean anything. It's not referring to the image as displayed by imshow(), since imshow() is not using the relevant color map to display the image.
A = imread('peppers.png');
However, imagesc() renders the image using the current colormap, and so the colorbar shows the mean pixel value. As the image is uint8 class, it's understandable that the mean values are somewhere in the midst of [0 255]
This might get a bit confusing as the behavior varies between RGB and single-channel images. For example, if you repeat the first example with a grayscale version of the image, then the colorbar does refer to the image pixel values.
In this case, it adopts a grayscale colormap, but there's nothing stopping you from using another colormap.