We're Eastern Europeans but his results showed 2.3% West Asian and 1.6% Inuit! How unusual is this?
According to my calculations, this means one of his 3x great grandparents was West Asian and going one generation back (4x great parents) was Inuit. Is that right?
No. Ethnicity allocations are at the rather more dubious end of DNA testing. The percentages quoted are a median that can vary widely (so for instance the quoted 1.6% Inuit could actually mean a range between a higher figure and zero - no match at all). They cannot be used to calculate a generational gap, and the testing companies say that their ethnicity allocations estimate genetic origins from at least 500 years ago or even more distant, so beyond the generational distance likely to be traceable in most family trees.
Because everything to do with ethnicity estimates is based on estimations and assumptions, some more loosely than others, whilst they can be considered fairly reliable at a continental level and with high probability percentages, very low percentages such as the two you quote should be treated with a great deal of caution and suspicion. So whilst they might genuinely indicate that your GF had some traces of ethnicity in common with those populations, they could be completely wrong I'm afraid, and the low percentages involved tend to make the latter more likely.
The percentages have nothing to do with a match length to another individual, they are averaged estimations of the amount of your GF's DNA that is believed may be identical to similar regions found to have commonality amongst the reference population samples that the testing company use to define their ethnicity regions.
The value of genealogical DNA testing is mainly in identifying matches between people who share identical segments of DNA that confirm they are related to each other by more recent ancestors; and above the lowest match lengths used by Ancestry and others, those matches can't be disputed.
Ethnicity estimates are a much looser guestimation but are commonly used as a carrot by the testing companies to attract consumer purchases by people who might have no or limited interest in genealogical research, and who may otherwise not have spent their money.
