CALO counters instability

There seems to still be some instability in the CALO counters. For instance, see:

https://lhcb-nightlies.web.cern.ch/logs/tests/nightly/lhcb-gaudi-head/2633/skylake_avx512+vecwid256-centos7-gcc9-opt/Moore/#all_94

Different content of Counters for algo FutureNeutralProtoPAlg
(FutureNeutralProtoPAlg ref) IsNotH for Pi0Merged | 326 | 122.1872 | 0.37481 | 0.29725 | 0.037118 | 0.99297
(FutureNeutralProtoPAlg new) IsNotH for Pi0Merged | 326 | 121.6586 | 0.37319 | 0.29670 | 0.037118 | 0.99297