Thank you for pointing that out. I have identified the type of situation that you describe in the diary review.
I understand that postal codes in Colombia are assigned to polygons that, naturally, cover the territory and thus, are overlaid on top of also streets, streets numbers, houses. However, as I summaries in the diary:
This is the dataset the Global Open Data Index must look for: a georeferenced geometry (in an official coordinate reference system) with main attributes - postcode and address information - street names, house numbers, housing complex. Because a postcode is most often assigned to a region within a city, this geometry should be polygonal.
I can see that in the dataset you have 2 attributes: "codigo" and "codigo_pos". If there would have been a dataset that made a connection between "codigo" and addresses (street name, number/ house number etc) at national level it would have been different.