With both bivariate and unclassed colors symbology, the current histogram-driven interface for viewing and adjusting class breaks is flawed in that when two breaks are close to one another, there is no way to select, modify, or even view the hidden number without modifying the other class. This can happen easily in datasets with a highly skewed normal distribution:
In this case, quantile class breaks resulted in two class breaks right around 30, but you wouldn't know it as there's a second number hiding under the 29.9 value. The only way to see what that hidden value is is to drag the other one away, at which point you're unable to go back to what you had without resetting the symbology for the entire layer.
Some symbology methods provide a different default, which is preferred - the table. The table view shows all class breaks at a glance and is much more easy to quickly than sliders -- I can just view or type the break value and move on. Please improve these symbology methods by giving the user the same fine-grain control found in other methods.