I don't think you and Dennis are really saying different things. Yes, you're correct in saying than an image is "just" data, but equally Dennis is correct in saying it's a special kind of organised data that facilitates (lossless) compression using algorithms specifically designed for the purpose. It does "make a difference" whether the data represents an image or not.
For example the lossless profile of the JPEG2000 standard is quite specifically designed for compressing 2D images; it will typically give much greater amounts of lossless compression (i.e. a smaller file size) than if you used, say, ZIP or LZW on the same data. Equally, the JPEG2000 compression algorithm will work very poorly, if at all, if applied to data that doesn't represent an image.
Richard.