"But ex post, holding the epidemic trajectory constant, it means that the disease is more transmissible per contact event than you expected."

...but that's wrong. The epidemic trajectory isn't constant (R0 is not an innate property of a virus; it changes with context), and it's pretty likely that we've completely mis-measured it by sampling from the wrong population. Our estimates for R0 are mostly model extrapolations from PCR testing data, which are not uniformly sampled from the population. If anything, in fact, they're heavily biased toward the susceptible population.

In other words, because we're sampling disproportionately from the most susceptible population, our estimates of R0 already seem high. If you sampled randomly from the population of less-susceptible people, you would see that the R0 is much lower than you originally predicted.

On top of that, there's absolutely no reason to believe that cellular immunity would make it seem like R0 is lower in any particular population: T-cell-mediated immunity is a late response. It can reduce the severity of an infection, but it's still highly likely you'd show up as a positive PCR test, and count toward an R0 estimate.

So you have a measurement based on a test that isn't strongly affected by cellular immunity, derived from a population biased toward the most susceptible individuals. If you have to doubt something here, you doubt the estimate of R0 (i.e. the model), not the observations of reality.