The extra function calls cost performance and in most cases the compared teams are NOT spectators, since most players are in existing teams {1,2,3}. Hence I tested the same functions excluding team=0 and my function resulted in a higher performance improvement (+40% vs +14%), because your function is only optimised for team=0 values.

function crosshairs.isfriendlyneutral(team1,team2)
	local product = team1 * team2
	--if product ~= 3 and product ~= 2 then
	if game("sv_gamemode")~= 1 and (product < 2 or product > 3) then	-- is not deathmatch
		return true
	end
end

Nice (though I feel like

game("sv_gamemode")

returns values as a string), but how is it better than this?

function crosshairs.isfriendlyneutral(team1,team2)
	if (team1 == 0 or team2 == 0) then
		return true
	end
	return ((tonumber(game("sv_gamemode")) ~= 1) and (math.min(team1,2) == math.min(team2,2))
end