An economically sustainable production of royal jelly (RJ) requires the use of genetics specifically improved for this very demanding trait. The French Royal Jelly Association has developed since 2010 a selection plan for that purpose, based on a collective data recording system and the initial seedstock of five French beekeepers. The aim of this study was to evaluate the efficiency of this collective breeding program, a decade after its concrete start. We considered three different genetic models to describe the phenotypes, hereafter named the Colony, Worker, and Queen Models. We showed that when the size and the structure of the dataset are insufficient to fit a Colony Model, fitting a Queen Model gives more understandable results for practical use of the estimated breeding values than the Worker Model. Regarding genetic parameters, we estimated moderate to high heritability values for all traits, regardless of the genetic model. Under a Queen Model, heritability estimates of RJ production were 0.19 ± 0.09 and 0.35 ± 0.13 for first and second harvests, respectively; the estimates were 0.76 ± 0.08, 0.29 ± 0.09, and 0.22 ± 0.08, respectively, for the calmness, gentleness, and spring development of the colony. We also proved that the breeding program is efficient, increasing the production at second harvest of 3.5% per year on average. Attention must, however, be paid to the evolution of inbreeding in this very small breeding population. Few genetic correlations were significantly different from zero between the traits, due to large standard errors of our estimates. It is, however, worth to note that the trends were generally favorable between all traits, which appears encouraging to improve production and behavioral traits jointly in the breeding program.