Update: This might interest you: On the Phonological Derivation and Behavior of Nasal Glides (R. L. Trigo Ferre, 1988). (PDF available here, I assume legally.)
I claim that nasal ‘absorption’ occurs when the occlusion of a nasal stop is removed or weakened considerably to the point where it is a glide. A nasal glide without any place features, [N], is shown to derive by a process which simultaneously reduces obstruent stops to glottal stops in Japanese, Choctaw and Malay.
I do not have the background to evaluate the whole of Trigo Ferre's argument, but my understanding from scanning the parts that relate to Japanese is that she would argue that the "y" you are hearing is the intervocalic /N/, or at least all that remains of it (apart from some vowel nasalization) in the word /seNeN/.
I am not an expert on Japanese phonetics, but this page has some data on similar words that suggests that nasalization is the key here (edit: which, of course, you did bring up; but I'm suggesting that it is the cause of what you're asking about).
- 真一 /siNiti/ → [ɕiĩtɕi]
- 本 /hoN/ → [hõ] or [hoɴ]
- 翻案 /hoNaN/ → [hoãɴ]
- 本意 /hoNi/ → [hõĩ]
- 電話 /deNwa/ → [deẽwa]
Maybe what you are hearing is not a glide, but rather the onset/release of nasalization (in [seẽɴ] or [seẽeɴ] or whatever).
Or, to restate sawa's suggestion in comments, it could be that the rule "/N/ turns into a nasalized vowel" is leading to a two-vowel sequence like [eẽ] or [ẽe], and people are inserting a glide to keep the two vowels apart (prevent "hiatus").
The difference between 千円 and 田園調布 could be due to accent as sawa says, or other factors (e.g. 千円 is arguably two words, 田園 is not; 千円 is very common, 田園 is not; etc.)
Incidentally, although I am not a native speaker... when I listen closely to myself saying 千円 what I hear is a narrowing (but not a stop!) in the uvular region, which is then released back into something more like a regular [e], and it is that release which sounds something like a "y". But it is definitely distinct from the palatal [j] in, say, 矢 /ya/.