One of the major dysfunctions of the science wars in the '90s was that people were using terms to yell at each other — social construction, relativism, postmodernism, paradigm, theory, science, fact, truth, reality — with very little in the way of clear and generally accepted definitions, even among partisans on one side of the dispute. Ian Hacking wrote a nice book on this problem, and the Wikipedia entry on "social constructionism" gives a nice summary of the argument in this introductory chapter. For your question, here's an especially relevant passage:
According to Hacking, "social construction" claims are not always clear about exactly what isn't "inevitable", or exactly what "should be done away with." Consider a hypothetical claim that quarks are "socially constructed". On one reading, this means that quarks themselves are not "inevitable" or "determined by the nature of things." On another reading, this means that our idea (or conceptualization, or understanding) of quarks is not "inevitable" or "determined by the nature of things".
The first reading treats social construction claims as ontological claims, about what kinds of things exist and why. Ontological readings of "the laws of physics are social conventions" might infer that gravity does not exist at all (in other words, it's a fiction), or it exists but only because social systems decree that it exists (like money).
The second reading treats social construction claims as epistemological claims, about our knowledge of the world. These readings would emphasize social contingency in the development of our knowledge. For example, the idea of laws of nature comes out of Abrahamic religion (Judaism, Christianity, and Islam): the laws of nature are rules laid down by God, the ultimate law-giver. If Christian-European societies hadn't dominated scientific development over the last 300 years, then we might not understand physics in terms of laws of nature. Similarly, a number of professional philosophers of science today (none of whom remotely resemble Lyotard or Irigaray) argue that the metaphor of laws of nature is counterproductive and misleading, even in physics, and suggest alternative conceptual frameworks for scientific knowledge. This wouldn't mean that we wouldn't fall if we walked out of Sokal's apartment window if China had dominated scientific development rather than Europe, or if we understood physics in terms of causal powers rather than laws. But it would mean that we would represent the phenomena of falling differently.
Sokal's criticism is only relevant to ontological readings of social construction, not epistemological readings.