Semantic features as the basis for semantic relations within the vocabulary: semantic markers vs. semantic distinguishers

As we know, the meaning of a lexical item is broken down into smaller sense components. The significant question in this connection is of course the following: how far the extraction of semantic components can or should be taken, or how 'delicate' the meaning analysis should be. Should we extract only those components (e.g. humanness, maleness, animateness, abstractness) which recur throughout the vocabulary of the language and which can plausibly be regarded as universal? Wouldn't this mean stopping short of the point at which all nonsynoymous words are assigned distinct componential analyses? There are non-recurrent, idiosyncratic features of meaning too