It is necessary to distinguish mutations from substitutions: assuming for simplicity that a mutation can happen at any place in the genetic sequence, most mutations are bound to result in non-functional genomes. Substitutions, on the other hand, are mutations that resulted in viable organisms - they can still have negative fitness effect, but not outright deleterious.
Note also that non-adaptive mutations are not necessarily bad - the neutral theory of evolution tells us that, due to randomness effects, substitutions may fix in the population without having a direct fitness advantage.
In viruses the quantity of substitutions can be very high - tens of percents, to the point that one has to introduce ad-hoc thresholds to distinguish viral strains, e.g., viruses that differ by 30% of their genetic content. How much of this is adaptive depends on the shape of the fitness landscape, as noted by @MaximilianPress. E.g., for HIV it has been found that up to a third of the sequence changes tend to revert to the ancestral HIV sequence, i.e., constitute host-specific adaptation.
In view of the discussion that followed, I would like to add some precisions to my answer:
- My distinction between mutations and substitutions is essentially distinction between the actual and the observed mutations. This is different from the more common usage, where substitution means point mutation, with mutation being a more general term.
- Since I mentioned the neutral theory, it is necessary to remark that the entirety of its claims is not generally accepted and/or supported by data. Moreover, its validity depends on the organisms in question. My use of the neutral theory is therefore limited to the fact that being adaptive/non-adaptive is not the only factor that determines that a mutation fixes in the population.