You're overthinking this. No need to worry about functional programming, purity, parallelisation, or any of that. The "map" concept in purely-functional programming languages is very interesting in its own right, but it has no more in common with Perl's map than mathematical functions have with Perl functions.
Perl's map is the wrong tool for the job purely because it does not do the job very well. It is good at iterating over a complete array and returning a new array, while the job was, IIUC, to iterate over part of an array and return a scalar.
If Perl's map did happen to do the job very well, then how could it be the wrong tool for it? For example, the pure FP map function must always produce exactly the same number of elements as it consumes. Perl's map can produce as many output elements as it likes for each input element, or it can skip it completely and produce nothing. In cases where that's desirable, it's perfectly appropriate to do it, without feeling guilty because a pure FP language would require an additional filtering step after the mapping was complete!