The present paper analyses the results of competitive blood-cell repopulation experiments in which Cx43-WT (connexin 43 wild-type) host mice, whose own HSCs (haemopoietic stem cells) were deleted, were grafted with fetal liver cells: 50% Gpi-1a (glucose phosphate isomerase-1a)/Cx43-WT cells competing with 50% Gpi-1b/Cx43-WT, 50% Gpi-1b/Cx43-HZ (heterozygous) or 50% Gpi-1b/Cx43-KO (knock-out) cells. The percentages of platelets, granulocytes, red cells, B-cells and T-cells containing Gpi-1b in blood samples obtained from 22 to 186 days after grafting, and the percentages of high-proliferation-potential colony-forming cells containing Gpi-1b at 255 days after grafting, were measured. The results show that, if we wait 4 months so that we measure the percentages of Gpi-1b end-cells formed by initially resting stem cells in the graft, values in HZ mice are greater than those in WT and KO mice by 10% or more. We propose a bipolar influence model for blood formation by grafted HSCs to explain this difference and other features of the data. Influence A is a direct one: for individual HSCs, the combined effect on HSC niching and HSC proliferation of Cx43 is superior to that of the KO allele. Influence B is a demographic one: HZ foundation mice compensate by having more HSCs than WT mice. The net outcome of influences A and B is that HZ is the winner.