Ah, this will be the issue then, and what you uploaded confirms it. Unlike our ncov workflow, the dengue workflow’s Snakefile doesn’t use a configuration file (builds.yaml). So your input data files (TSV and FASTA) and subsampling schemes aren’t being used by the workflow.
I expect the Colombian sequences you found in the tree you generated are from NCBI GenBank, by way of our curated dengue data files that are downloaded at the start of the workflow.
The dengue workflow currently isn’t set up well to accept custom input files or custom subsampling schemes, although it’s on our list of things to do.