What are the patterns exactly

Being confined to the subscales that look like pattern-matching and analogies, IIRC; I'm not sure which paper I get this from, but it seems Jensen does at least make this claim in http://www.charlesdarwinresearch.org/2010%20Editorial%20for%20Intelligence.pdf and in some citations in http://menghusblog.wordpress.com/2013/06/21/explanation-behind-the-non-g-gains-in-the-flynn-effect-introducing-the-measurement-invariance-model/ so there's some starting points at least.

how do we know what inbreeding depression costs? From recent studies of inbred individuals?

Yes, that's how one would do it. The usual reference is to a study of Japanese cousin-marriages back in the '50s or so where IIRC the estimate was <5 IQ points, but there's been research since then, of course; a google for 'inbreeding depression intelligence' should bring some research to light.