I am dubious of this assertion. The original 1960 paper "Recursive Functions of Symbolic Expressions and Their Computation by Machine" introduces S-expressions and M-expressions as explicit entities (not as some abstract syntax, but as a literal syntax that people were able to write down). Now, M-expressions never really got used because everyone just used S-expressions, since they were simpler.
Do you have some citation about how he left implementation of syntax as an exercise? I'd be interested to see that.
I think he's kind of close, at least in that it was not meant to be written as S-expressions. AFAIK, the M-expressions were meant to be the main syntax. The very early Lisp docs I've seen were all written in M-expressions, with S-expressions taking the lead something like a decade later. (This was all decades before my time, so it's possible this is sampling bias, but I don't think so.)
Do you have some citation about how he left implementation of syntax as an exercise? I'd be interested to see that.