I know that compound words are made up with two small words, but is "tax-free" or "timeless" compound word? How about "thought-free"?
A lot depends on your theory of morphology - see e.g. Lieber and Štekauer 2011 - see esp. 1.1.4 Summary. Several tests for compounding have been proposed; the biggest problem is that they do not necessarily yield the same results. And then there is no clear-cut, universal boundary between a free word-form and a bound affix. As to be expected, by its very nature, no single theory can encompass all language data into its procrustean bed.
That being said, Bauer, Lieber and Plag 2013, in The Oxford Reference Guide to English Morphology, write the following:
"The semantically related formations involving -free and ill- are sometimes treated as affixes or affixoids, but we consider formations with these elements as compounds, on the grounds that there are no compelling arguments for their status as affixes or affixoids (see Chapter 16 and Dalton-Puffer and Plag (2000) for pertinent tests" (p. 354).
This question is likely based on the assumption that less is a proper word that can compound, and thus drrives the morpheme. At least, that was my first impulse to counter @BillJ's comment.
However, similar to most and -most, the morphemes are likely not the same, though historically entangled (cf. wiki/-most).
From Middle English -les, -leas, from Old English -lēas (“-less”), from lēas (“devoid of, loose from, false”), from Proto-Germanic *lausaz (“loose”). (Not related to less, which derives from *laisiz, *laisizô.)
[wiki/-less; emphasis mine).
I wouldn't say its not related at all, if a plausible connection might be too distant. But note the different umlaut anyway: *laus- and *lais-.
It figures that the pronounciation differs, to say the least: whereas -free has stress, that's not the case for -less. The vowel therefore differs from that in less.
Hence, it's debatable whether language acquisition has them as one stem, or if the association, that invariably exists, is more difficult than straight up derivation would be.
Point in case, the written form likely confuses us. At least a second language speaker learning reading and listening in parallel will be confused by a comparison with -ful--with a single l--that I see now however in a difficult cluster with German gefallen, Gefühl, but not--to be contrarian--not with full, as if full of awe. It's more complicated than that, and certainly we can observe that -ly as a suffix would be there for a learner--not so much for the current historian--to consider for comparison.
All that doesn't exclude that some speakers have this, who are not bound by historic analysis, and operate from a less than absolute synchronic analysis. But only in English, whereas German has e.g. arbeitslos (workless, "unemployed"), where los rather seems to correspond to adj. lose "loose", not "lose"; or to los, was ist los? "what's going off?", los gehen, "let's go" (a radio dj once mistranslated "equal it goes loose", i.e. gleich geht's los).