True, but randomly occuring mutations (random with respect to the selection function) will yield a system that moves with jumps of all sizes.
There is also the genotype-phenotype relationship. It may take several small changes in genotype to make any change at all in phenotype. Likewise, some small genotypical changes may result in large phenotypical change.
I think that "random" here would be shorthand for "mutation not necessarily related to selection."
The more I think about this, the less impressed I am by the randomness of mutations. It seems that mutations are very common. I've read that each human conceived may have 100 mutations (of various types, most irrelevant). Given that kind of genetic background, we could almost consider mutations as a given, much as we assume that the origin of life is a given. By that I mean that it almost doesn't matter where life (or mutations) come from. They're here! That said, natural selection is the whole game. And that's not random. But I suppose mine is a fringe view. No big deal.
True, but for several reasons mutations of large "size" (i.e., a large effect on the phenotype) will be far more likely to be harmful (as well as immediately fatal) than small changes.
In fact, for a continuous fitness function it can be shown mathematically that as the size of the change approaches zero (i.e., for smaller and smaller changes), the odds of a mutation being beneficial (albeit to a correspondingly small degree) approach 50%.
Dawkins discusses this point in a few of his books (including "The Blind Watchmaker" and "The Ancestor's Tale"), giving credit to statistician and biologist R. A. Fisher for having made the point originally. A direct consequence of this observation is that evolution is far more likely to proceed via accumulated small changes than by mutational "leaps" of larger effect.