All three of the records are on the GRO indexes (in various name combinations), and I have sent the information to Chris via PM.
In this case, I think the most likely scenario is that the children were originally registered as being to a married couple - but I don't think the parents were in fact ever married. That allowed a later correction to be made to each entry which would have the effect of removing the father's details ( as he probably wasn't present when the births were registered), but the correction information would still appear on any certificate produced (even though his name would be crossed out), so later on the three births were re-registered to have the effect of producing a register entry without his name at all.
But that is a possible answer - without the certificates from all three registrations, and three later re-registrations, it is impossible to be certain.
A very good example of how it is very difficult to know exactly what surnames are on a birth register entry, whether a father is named, or even if the parents are married or not just from the information in the indexes alone. Comparing how entries appear in the old and new indexes can be useful if they differ, but you do have to get the certificates otherwise it is always assumption and guesswork.