Basically it might be possible that you’re using a different “Alive Test” method (as described previously) during the “full” scan then the default which ping_host.nasl is using (ICMP Echo request) when running via openvas-nasl.
As VTs running via openvas-nasl are not able to read any of the script_add_preference() set in the metadata of a VT or within a scan config you might need to play around with the script_get_preference calls in ping_host.nasl to get a similar behavior like in a “full” scan.
Generally please note that openvas-nasl is not really a replacement for a “full” scan via openvassd and should be only seen as a development tool for the initial development of a VT.
In general it doesn’t care if a host is “dead” or not so the ping_host.nasl dependency is unnecessary when using openvas-nasl.
Furthermore it doesn’t evaluate most of the data in the description block like e.g. script_require_ports, script_dependencies, script_mandatory_keys or the previously mentioned script_add_preference.
Some of these points are also described in Understanding testing of nasl scripts, especially the note on unscanned_closed and unscanned_closed_udp mentioned there might be relevant as well.