regex – Regexp to remove small numbers and leave large ones

On text files with many paragraphs, sentences or phrases are numbered.  I use regexp in perl to remove those numbers.  They are always followed by the first letter of the sentence/phrase or by a space.  But that also matches numbers legitimately part of the text.  If I could limit it to a string of one or two digits, not more, which does not contain a comma, I could manually delete the rare instances of an unwanted three-digit number, or re-insert the rare two-digit number that shouldn’t have been deleted.

I haven’t been able to figure out a regexp with those limitations.  How can that be done?

Example:

perl -p -i -e 's:(\D)\d{1,2}(\w):\1\2:g;
               s:\d+-\d+::g;
               s:^\d{1,2} ?::g;' {filenames}

removed the markers but also removed the digits from “337,000” leaving the comma.