African swine fever (ASF) has become the major threat to the global swine industry. Lack of available commercial vaccines complicates the implementation of global control strategies. So far, only live attenuated ASF viruses (ASFV) have demonstrated solid protection efficacy at the experimental level. The implementation of molecular techniques has allowed the generation of a collection of deletion mutants lacking ASFV-specific virulence factors, some of them with promising potential as vaccine candidates against the pandemic genotype II ASFV strain currently circulating in Africa, Europe, Asia and Oceania. Despite promising results, there is room for improvement, mainly from the biosafety point of view. Aiming to improve the safety of BA71∆CD2, a cross-protective recombinant live attenuated virus (LAV) lacking the ASFV CD2v gene (encoding β-glucuronidase as a reporter gene) available in our laboratory, three new recombinants were generated using BA71∆CD2 as a template: the single mutant BA71∆CD2 f, this time containing the fluorescent mCherry reporter gene instead of CD2v, and two double recombinants lacking CD2v and either the lectin gene (EP153R) or the uridine kinase (UK) gene (DP96R). Comparative in vivo experiments using BA71∆CD2 f, BA71∆CD2DP96R and BA71∆CD2EP153R recombinant viruses as immunogens, demonstrated that deletion of either DP96R or EP153R from BA71∆CD2 f decreases vaccine efficacy and does not improve safety. Our results additionally confirm ASFV challenge as the only available method today to evaluate the protective efficacy of any experimental vaccine. We believe that understanding the fine equilibrium between attenuation and inducing protection in vivo deserves further study and might contribute to more rational vaccine designs in the future.