When looking at casualty statistics for WWI generally, but especially in France, it is important to remember that desperately haphazard mobilization efforts, the astonishingly concentrated nature of much of the mortality, and a zealous culture of hiding casualty statistics from the government much less the public mean that the dry statistical tables that you will find hide a lot of uncertainty. In a unit that is missing a third of its men, who counts the bodies? Medical units also only receive the wounded and cadavers who can be practically carried back, and so only receive a portion affected by biases that can't be practically accounted for. So the best way to figure out who died that the High Command arrived at early on was to do their best to have some idea as to how many soldiers were active in each unit at any given time, attempt to count who showed up where they belonged after battles, attempt to count who showed up to medical units, and then to make maximally informed guesses based on the difference. This rough methodology naturally leaves a lot of room for double and undercounting on practically every level, but also very little room for comprehensively understanding very much about who died aside from often rank, division, and combat arm as well as what can be inferred from them.

That said, approximately 1,397,800 French soldiers died in uniform during The Great War out of a total population of 39.6 million^([1]). This would be a particularly extreme selection event when you consider the dead represent a horrifyingly high proportion of the male half of a relatively narrow generation of this total figure. Thus, at first blush, it doesn't seem so implausible that odd selection biases in these deaths could have noticeable impacts on who you would see walking around in France after the war compared to before. However, I'm not sure that we would need a direct accounting of the relationship between hight and war mortality in France to discount this theory. At least as I understand it, the idea is that taller men would be more vulnerable to having their heads exposed over a parapet while manning a forward trench, and thus more likely to die by a sniper's bullet right?

While the threat, mortality, and morbidity imposed by snipers clearly had a profoundly outsized impact on morale, experience, and memory of the war, it was also clearly a remarkably small part of what actually caused casualties. For the most part, WWI was a war primarily fought by artillery, particularly on the Western Front. For example, field and heavy artillery caused 60% of French war wounds^([2]), though importantly this will relatively overrepresent those hit by artillery near logistical centers while underrepresenting those shot, stabbed, or burned away from them, and also doesn't account for likely large differences in the survivability of these very different kinds of injuries. The vast majority of other deaths would be due to illness or during major engagements.

For the most part, rotating through a forward trench on the Western Front for a period of time meters away from the enemy was a remarkably safe thing to do given the circumstances. The men in the opposing forward trench would be among the least of the threats to your health compared to the heavy artillery behind them, the mud, the fecal-oral pathogens in that mud, and the trauma imposed by the context. Survivor accounts suggest that activities like sniping and trench raids would have a way of reciprocally dissipating over time as the threat of reprisal deterred them, and had to be creatively encouraged to continue.

Thus, the real threats to mortality for French soldiers likely did not meaningfully select for hight.

^([1]) Huber, Michel. La Population de la France pendant la Guerre, Paris 1931.

^([2]) Ministère de la Guerre, Direction du Service de Santé, Etude de statistique chirurgicale, Guerre de 1914-1918, Paris 1924.

EDIT: I have found the following paper: "Big and tall soldiers are more likely to survive battle: a possible explanation for the ‘returning soldier effect’ on the secondary sex ratio" but have not included it as I am not especially convinced by the statistical case it makes. More modern machine reading techniques than were necessarily accessible at the time applied to the data set could definitely do some cool things.