I have tried on latest Ubuntu beta and CPU metrics are properly collected, so I don't think it is related to this. Also mention that both, the graph and the gauge use the same data, so if it appears in one of the visualizations then it must be properly collected and stored.
The main difference I see is that the graphs show any point stored in Elasticsearch as is, it doesn't matter if there are gaps in this data. But the gauge does some calculation and is intended to show in a glance the current status of the visualized metrics, so a gauge is more useful with a good, continuous dataset and specially with data in the final time range.
To better illustrate this, look at this example, it shows the same dataset, with different time ranges, the first one with metrics in most of the range:
And the second one after a while without metrics:
Are you having problems after collecting data for a while and visualizing a time range with data in all the period (as in the first example)?