The concepts dominance, additivity, epistasis, recessivity and others are all specific to a given phenotype. Here you are getting confused because you are considering two different phenotypes at once.
Slightly longer answer
By definition, if an allele is recessive, then in the heterozygote state, the phenotype of interest is just like the homozygous dominant. An important point in the above sentence is "the phenotype of interest". A given locus may show a pattern of simple dominance-recessivity for a given phenotype but may show a completely different pattern for another phenotypic trait.
Let's consider your example of sickle cell anemia. For the phenotype which is the disease, sickle cell anemia has a simple dominance-recessivity relationship. However, for the phenotypic trait which is fitness (yes, in quantitative genetics, fitness is often modelled as a simple phenotypic trait), sickle-cell anemia shows (environment dependent) heterozygote advantage. Also, the fitnesses of the two homozygotes differ greatly.
Note that, in reality, very few loci show cases of perfect recessivity dominance.
You are considering two (or even three) phenotypes.
First Phenotype: Fitness
Fitness shows overdominance (with highly unequal fitness for both homozygotes).
Second phenotype: Disease
Double mutant are sick, the others aren't. Perfect dominance-recessivity (although if I am not mistaken heterozygotes individuals have some restriction when it comes to scuba diving if I am not mistaken)
Third phenotype: Shape of hemoglobin
If the heterozygote hemoglobin is shaped just like the healthy homozygote, then there is dominance-recessivity. If the shape differs (which might be causing malaria resistance, I don't know), then there is some partial dominance going on.