Since you are comparing two classified images with a small range of values, instead of using the image analysis window functions (especially given the cryptic result), why not just use built in spatial analyst functions?

There are a number of ways to do this, but I suggest using Combine: This will output a new grid with a single value for every unique combination of input values, e.g. forest-forest, pasture-forest, forest-pasture, pasture-pasture, etc. The output attribute table will show all original and new values, allowing you to assign new names to the mapping (e.g. deforestation, regrowth, no change, etc.).

Since you are comparing two classified images with a small range of values, instead of using the image analysis window functions (especially given the cryptic result), why not just use spatial analyst functions?

There are a number of ways to do this, but I suggest using Combine: This will output a new grid with a single value for every unique combination of input values, e.g. forest-forest, pasture-forest, forest-pasture, pasture-pasture, etc. The output attribute table will show all original and new values, allowing you to assign new names to the mapping (e.g. deforestation, regrowth, no change, etc.).