I don't find that surprising at all, achievement stats don't take owned copies vs played copies into account. And many people probably played the game on GOG version too...
So generally speaking, I think the number is more like 10% to 20%
I think this reasoning is flawed. Not everybody plays this game on Steam; that much is true. But why do you assume the people playing it on Steam are different from the people who do not? The Steam achievement says that 5% of the Steam-players have completed the game. It does not say anything of the people that do not play it on Steam, but why assume that they play it differently?
Let us assume that 50% of the players play it on Steam and 50% do not. To make your assumption true - using the completion rate of 10% - the number of non-Steam players that have to complete the game must be 15%. If the upper rate of 20% is to be true, the non-Steam player completion rate has to be 35%. This is a rather large difference of the two groups. Why do you assume they are so different?
I do not know how big the different groups are. How many play this game on Steam and how many do not. If for instance only 1% of the players play it on Steam and 99% of the players do not, the numbers provided by Steam cannot be trusted to apply for the entire gaming base. If the groups are almost equal or if the Steam group is the bigger group, I think the numbers provided by Steam are quite accurate. I cannot see why the behavior of the two groups should differ that much with regard to completing the game.
I know this is a bit of a derail, but the "scientist" inside me had to speak up