Should I set normalization factors to 1 if it's expected that most DEGs are upregulated or should I calculate them using TMM using calcNormFactors?

There is no problem with most DEGs being differential in the same direction, providing that the total proportion of DEGs is no more than about 30% of all genes. You can perform a standard DE analysis including TMM normalization.

It would certainly be wrong to set normalization factors to 1. That is never correct, but especially not when the DEGs are unidirectional. When there is a large amount of unidirectional DE, the normalization factors should always be less than 1 in the condition with the upregulation.