Limitations of scanned human copresence encounters for modelling proximity-borne malware