IPA
It’s a well-established principle when using IPA for phonetic descriptions to declutter the transcription by leaving out anything that can be predicted using knowledge of the language. (“It is better to state such information in the conventions that accompany a phonetic transcription rather than in the transcription itself.”)
For instance, when giving the original pronunciation of a borrowing in a monolingual dictionary, the IPA is often simply copied from a bilingual dictionary. That is, the pronunciation of ‘Champs-Elysées’ might be given as /???z elize/, which is fine if you know French IPA conventions, but not otherwise. Where does the stress go – equally on each syllable, or will one syllable get more stress? Is the /e/ to be pronounced [?] as in English? Is the /l/ light or dark?
Or going the other other way, if the word ‘tell’ was taken into French, giving it pronunciation as /tel/ would indicate [t?el] rather than [t???].
Surely in this context it would be better to use precise IPA.
Surely in this context it would be better to use a proper phonetician to supply the transcriptions, and do so in keeping with the overall design of the transcriptions used in a certain title.
For instance, [bʌtən], [bʌtn] and [bʌtⁿ] may all be perfectly good transcriptions of “button” in the abstract, but you can’t know what the right one is in context without knowing how parallel cases like batten, battle and batter have been handled.
Unfortunately in my experience most lexicographers know little about phonetics and care less. They assume a transcription is just a transcription and can be copied and pasted at will, or just chucked in off the top of the head of someone who was vaguely familiar with IPA at university 20 years ago. It doesn’t occur to them that a dictionary has style in the matter of phonetic transcriptions just as much as in the use of bold or italic, and you need to know what you’re doing.
However, with regard to your main point, the general principle you mention still holds as far as I can see. The question of how stress, velarisation etc works in the language should be outlined in the front matter where the transcription is explained. Also, there is the principle of simplicity, and perhaps one of inclusivity as well: the narrower and fussier you make your transcriptions the harder they are to read (and typeset), and the more rigidly they impose just one accent of English. At “battle”, something like /batl/ would probably work fine anywhere in the world. Imagine the hassle, and typographical complexity, of making that specific to, say, RP ([æ], lateral release, dark l) as opposed to GenAm ([ɛ~a]?, some nasalisation in the vowel perhaps, t-voicing almost but not quite always) or African ([ɛ], central release (perhaps even aspirated?), light l) — and for what benefit?
What would be useful would be to put more thought into the way words and even phrases are typically reduced in casual speech. Words like “and” or “don’t” only sometimes have their citation form, and might come out as [æn, ən, n], [dəUn, dən, dn] etc, not to mention phrases like [gənə] for “going to” which are far more common than their written version if any (“gonna”); we nearly always elide the first vowel in suppose, but would rarely write “spose” except to make a special point of it.
I imagine these reduced forms are quite problematic to a learner, and not necssarily predictable from general principles, while using the careful forms [ænd] and [dəUnt] all the time will make them sound distinctly pedantic and foreign. Good dictionaries make some attempt in this direction but not enough.
PS: I’d quibble with your interpretation of English /l/ in [tʰɛʟ]. Perhaps this is just a problem with how the symbols appear in my browser, but as I’m sure you know upper-case L represents a velar, rather than velarised l. As an English native speaker I’m not sure I can even pronounce [ʟ]!