Thanks for raising this.
When you click through to the LGA page, the first tab ranks all suburbs within the LGA. The data in every row of the table is referent to each suburb listed and not to the LGA as a whole.
You did correctly point out that there is only one locality in this LGA, and the expectation would be that the LGA values should be comparable to the locality values in the table (Kalbarri). However, we calculate the LGA-level metrics independently from suburb/locality -level metrics.
As we only list the suburbs/localities for which there is significant sales data and the rest of the suburbs are omitted, it is possible for aggregated LGA values to differ from each individual locality within the LGA. In other words, even though the data from other localities in the LGA (15 in the case of Northampton) is accounted for in the calculation of LGA-level metrics, it is so thinly spread across multiple localities that it only transpires in 1 locality out of 15 at the suburb-level.
Also note the Very Low confidence assigned to Kalbarri, indicating the low sales volume and therefore high data error in this particular locality.
This scenario is more of an exception than a rule as you have come across a rural area with very limited data. If you are interested in researching this area, I would suggest focusing on the LGA-level data and discarding suburb-level data due to Very Low Confidence ranking.