I think we're seeing more pixellation with one numeral than the others because a multi-layered PDF document has been optimized without first flattening the layers. Though this seems to be done as a single operation, in reality each layer is optimized sequentially for greatest visual clarity independent of the other layers. When the software optimizes a layer with a lot of visual data—many characters or numerals—the overall effect is different than when the same process is applied to a layer containing a single or very few characters.