The mass distribution isn't necessarily a good proxy for the deflection. Lyman alpha lines are from absorption by hydrogen gas clouds; the average effect of those on the path of light is likely to be close to zero, because the clouds are diffuse and more or less homogeneous (i.e., their average density on large length scales is pretty much the same everywhere). The "Tardis region" model requires large deflections of light at the boundaries between the "Tardis" regions and regions of normal density, which will not average out since there aren't enough such boundaries between us and the distant supernovas that are our main line of evidence for the accelerating expansion of the universe.