Now I can confirm that you can create unique female lines in dlg if you just make a copy of male mp3 and vcd file and just rename it to *_col_f.*
Then it plays for female character.
The one *_col_f.vcd file with link to *_col_e.mp3 (or any other required to be played) inside will be quite enough, and you don't need to replicate .lips and audio with same names, because DLG mechanism works to refer certain dialog line to .vcd file first (and audio-links inside it), and ONLY if it's missing, it seeks for .wav or .mp3s. Lip files in fact are just "postfixes" to existing sound files to use extra data (in HL2 were merged with .wavs at all), and they will never be played if the same-named audio not found (no matter, is vcd file here or not).
p.s. If you working on lips yourself, please check the new SDK update, where new tools for localization and making lipsync data included. Now it's easier than whenever :)