I think John Lawler and others make a good point in that "antonyms" are vague, and I suspect that, despite the descriptivist intent, the question arises from a semantic issue.
From Wiktionary, an antonym is "a word which has the opposite meaning of another, although not necessarily in all its senses." Thus fast is an antonym of slow, but fast is also an antonym of eat. However, most of us wouldn't think about comparing speed with consumption. Useful can be interpreted as "having non-zero utility," which means the opposite of useless. However, useful can also mean "having a positive degree of utility" which is not the opposite of useless. So they are fine antonyms, but not opposite in all meanings. A more appropriate opposite for the comparative version of useful would be harmful or detrimental.
For the more descriptive questions, specifically regarding the "-ful" and "-less" suffixes, I suspect that use of these words depend on how these suffixes are commonly interpreted. "Doubtless" and "useless," for example, imply devoid of doubt and devoid of use. "Thoughtless" and "tasteless," for example, imply lacking thought and lacking taste. The latter pair would be more common in comparative relative to non-comparative use since one can be naturally seen as more or less lacking. The former pair is less commonly seen since it is less logical and descriptively less common (though not unthinkable) to be seen as more or less devoid (of course, cf. emptiest). In general the commonality of use seems to me in line with whether or not it is logical -- so I don't see them as necessarily in conflict.
However, one exception comes to my mind (not saying that there aren't others). When raukh mentioned "impossible" (p = 0), my first thought of an antonym was "certain" (p = 1). As someone more accustomed to speaking with statisticians, for me, it sounds awkward when someone says something is more or less certain. However, I recognize that both descriptively and formally, certain is a comparative adjective. Indeed, it seems that the use of certain as a comparative is more common than the use of uncertain as a comparative, although that appears to be in relative decline.
Additionally -- this is perhaps silly of me to think it needs stating -- choice of which words to use also depend upon the emphasis of the sentence, even for paired words. Whether someting is "more impossible" or "less possible" may, for some, have different connotations. Curiously, those words seem to be converging in frequency of use.
Best Answer
Preamble on the logical implication of an antonym
You are trying to confuse people's understanding of the term antonym.
Logically, speaking, the term you could have used is antithesis. However, since all the dictionaries define antithesis as exact/direct opposite, you should use the mathematically more precise term of null-hypothesis
The antonym of good = bad
The null-hypothesis of good = not good <= {bad, not bad but not good}
For the hypothesis
John = good person
then the null-hypothesis of the correlation is
John = anything but good.
In English, it is very frequent for an antonym not being the exact null-hypothesis. In fact, it is frequently true that almost all natural languages have antonyms not being the null-hypothesis. This is due to the absence of a pure binary set of possible states in such cases.
If a phenomenon Z has the possible states sorted by intensity {a, b, c, d}, then we can be sure that
antonym of a = d
But, the null-hypothesis of a = not a = {b, c, d}
What is more difficult when the states of a phenomenon is unsortable or unrankable, then any state within such a phenomenon does not have a generally acceptable antonym.
Therefore, the antonym
Therefore the antonym of the word prefer is indeed among words describing its extreme opposite
You are asking for the null-hypothesis of prefer, and as I have anecdotally stated, the null-hypothesis of many words is not the generally accepted antonym, due to having non-binary repertoire of states. Prefer is one of those words.