We have a search for forms and publications feature on a wide range of subjects on a government site for our users. We're trying to develop a filter option that will narrow the results based on a subject (e.g. Electrical). The content producers of the forms and publications have proposed subjects that matches the information architecture of the site pages. As a UX/IA designer, this doesn't seem the right approach given the changing nature of the site architecture, and some proposed subjects would be too-specific, resulting in few results. I need to provide qualified reasons to reject their proposed taxonomy. Are there resources from studies that could help?
Sometimes you don't need 'qualified' research to justify a decision that is based on 'common sense' logic.
If as you say, the filter is going to be based on the IA of the site pages and the site architecture will change with some level of frequency, then the question to be asked is whether the filter option will need to be adjusted to consistently match this or not. If not then the original rationale for the design of the filter is then invalid (or needs to be reconsidered).
In terms of the proposed subjects being too-specific and resulting in few results, that's not necessarily a bad thing to reduce the amount of results returned, but there are other considerations such as whether the same IA would result in users not being able to locate certain information on the site as well.
So I think it is worth calling out the decisions being made and the rationale/logic behind it. In my opinion it feels contradictory and is likely to cause issues down the track when you have to make other decisions.