it first copies character's pixels to an area of memory until the word or phrase is finished.
This is probably your first chance to make it copy 2 pxl lower.
Once the phrase is complete, the entire section of pixels is then copied to another place in memory as a sprite (probably when looking at the memory dump)
And this is your second chance.
Find the images quickly is a matter of fantasy first (while reading the names of the files), logic (analizing the (tree) structure of the game), observation (a menu graphic will probably be 2/4 bpp but not 8 opposite to a "complex" background image will probably be 4/8 bpp and not 2bpp), observation 2 (depending of the bpp, you have another size of palette), knowledge ( being capable to use CT2 well and understand what each command means), knowledge 2 (know the image formats
used by the NDS), intelligence (try to view a graphic when CT2 have a book icon in front of that file with the letters "LZ" is very stupid) and experience (by the look at the hex code, already have an idea of what you are looking at).