You should aim to model a neighborhood with a diameter of around 100m. The building energy calculation of Dragonfly’s UWG component is based on the Building Energy Model Town Energy Balance (BEM-TEB) scheme, which can model the building energy consumption of a city (approx. 1000m to 10,000m diameter) from the resolution of a neighborhood (approx. 100m). From this neighborhood model, the UWG component will derive a representative urban canyon with geometric dimensions reflecting the weighted average of your neighborhood typologies.
However, that neighborhood has to represent a reasonable city typology. The UWG uses a lot of microclimate, and mesoscale parameters based on this assumption. If you model a neighborhood that is more suburban, or rural in nature (i.e very low site coverage, density, and high amounts of vegetation) the UHI results are going to start being more and more incorrect.
I’m not sure how you are representing that extra site area (vegetation? roads?) but I think you need to rethink your assumptions. If you want to test the UHI impact of the 15 x 15 versus 25 x 25 typology, then you should just build an entire neighborhood of just those typologies, that is then extrapolated to a larger scale to derive the potential UHI impact. If there is a large non-built site area in your neighborhood, you can incorporate that into your model, but just be aware the probability of error is higher, and the larger site area is going to dominate your microclimate interactions and wash out differences between your two different typologies.