See also s.v. lšm prep., a usage often difficult to distinguish semantically from l_ prep. preceding this word.
We assume that in JLAtg there was at one time a distinction of a sort (i.e., idiolectal and/or between absolute and construct forms) among the three attested forms with short /o/,/u/, and /e/ vowels, but the versions of the texts at our disposal do not provide any clear evidence.