Typical Qualitative Risk Assessment Descriptors Don't Work

You’re probably familiar with the risk evaluation matrices that require you to rank severity of harm and multiply it by the likelihood of that harm being realised.
The ranking of severity usually poses no major problems, eg:
5 - Catastrophic loss of life
4 - Fatality
3 - Life changing injury/illness
2 - Lost time injury/illness
1 - Minor injury/illness
The ranking of likelihood is a bit more challenging though, eg:
5 - Very likely
4 - Likely
3 - Probable
2 - Unlikely
1 - Very unlikely
Is everyone going to be on the same page when it comes to the difference between ‘probable’ and ‘unlikely’, or, for that matter 'very unlikely’ and simply 'unlikely’?