One book says...
The general rule of using the dual number system btwn 1 & 10 are; 1. Use Japanese forms with words of Japanese origin. 2. Use Chinese forms with words of Chinese origin. A natural result of not having aboriginal Japanese numerals beyond 10 is that the Chinese form is the only choice. The problem is the Japanese vs. Chinese distinction seems to have disappeared in the three numerals 4 Jp./yon/ ~Ch./shi/, 7 Jp./nana/ ~Ch./shichi/, and 9 Ch.G./ku/ ~Ch.K./kyuu/. In combination with counter nouns -mai(sheet), -dai(set), -ko(piece), -hon(stick), -satsu(volume), -hiki(insects, animals), -too(head), -wa(birds, rabbits), -nin(person), & -nichi(day); Ch. /shi/ 4 is not used at all; Jp. /nana/ 7 seems to have taken over Ch. /shichi/; and kanon /kyuu/ is preferred over go'on /ku/ in most cases with the following exceptions.
1. 7 /shichi/, 17 /juu-shichi/ are used interchangeably with /nana/ /juu-nana/: -mai
2. 7 /shichi/ is used primarily but interchangeably with /nana/: -wa, -nin
3. 17 /juu-shichi/ is used exclusively: -nichi
4. 9 /ku/, 19 /juu-ku/ are used interchangeably with /kyuu/, /juu-kyuu/ -mai
5. 9 /ku/, 19 /juu-ku/ are used primarily but interchangeably with /kyuu/, /juu-kyuu/: -nin
6. 19 /juu-ku/ is used exclusively: -nichi
I don't have a table for ナ。ツ「 , but the general preference for Jp. /nana/ 7 and kanon /kyuu/ 9 in combination with these counter words does not contradict your observation. (Pimsleur's usage in the early chapters may be based on a broader base of examples, maybe?) Reading of 17 ナ。ツ「 as /juu-shichi-en/ would have to fall in class 1. of exceptions; but not 4 or 5. These exceptions appear to be random, depending on general usage. I'd like to hear of a simpler explanation, though. Furthermore, the language of reference is that of the late 1950's to early 1960's, Tokyo dialect. Any changes in the past half century?
source: 窶愿コ窶怒ナ津ェナスlツ週ナ?テ? Yoshio Ogawa, Daigakusyorin, 1963, Tokyo.