A red flower does not have to turn white to lose information in a photograph. The red component can reach a representation's upper limit while its green and blue components remain lower. Subtle differences between petals may then be flattened even though the picture as a whole looks comfortably exposed. That is why checking brightness alone can miss a color problem.
A camera already makes decisions about the color of the illumination, and those decisions affect the rendered RGB values you inspect. Try rebuilding a target in Color Matching for an informal feel for how color and lightness can vary separately. The game is not an exposure meter, but it makes the distinction less abstract.
Three limits behind one bright picture
An RGB image describes each pixel using red, green, and blue components. The values are coordinates in an image representation, not a direct report of three pure wavelengths arriving from the flower. A rendered red petal may have a large red value and much smaller green and blue values. Increasing its brightness can push the largest component toward the limit before the others get there.
For an explicitly simplified eight-bit example, suppose two neighboring pixels would be represented as red values of 260 and 275 after an adjustment. That representation stops at 255, so both become 255 in that component. The numbers are an illustration, not a measurement of a camera. They show why turning the finished image down later cannot reveal which of the two red values was originally larger.
A neutral bright patch can push several components toward their limits together. A saturated colored object can stress them unequally. Thus a red petal, a blue stage light, and a white window are different tests of the pipeline. None should be judged solely by whether the final pixel is white.
Read the colored graphs as distributions
Nikon's camera manual describes separate red, green, and blue histograms. Across a histogram, the horizontal position represents the component's brightness value and the height represents how many pixels have that value. It is a distribution, not a map of where those pixels sit in the frame. A large spike may come from one broad uniform area or from many scattered areas.
Choose the RGB display when photographing strongly colored subjects. Look for a component crowding the right edge, then inspect the subject itself at an appropriate magnification. A graph touching the endpoint is a reason to investigate rather than automatic proof that the picture is unacceptable. Some scenes intentionally contain maximum values, and the camera's histogram is a guide to its image rendering.
That last distinction matters. Nikon explicitly notes that camera histograms are only guides and can differ from those shown in imaging software. Do not treat the playback graph as a calibrated diagram of unprocessed sensor capacity. It is useful evidence about the picture the camera is showing you, with settings and processing already involved.
The clipping may happen after capture
There are at least two places to lose the distinction. At capture, a sensor's finite capacity can be exceeded. Later, processing can push values beyond the range of a chosen output. These are related limits, but they are not interchangeable. A rendered preview can look clipped while a raw workflow still has useful information to produce a different rendition.
Adobe's Camera Raw documentation describes clipping in terms of values that cannot be represented and notes that the histogram changes as adjustments change. This gives you a practical diagnostic: return exposure and color adjustments toward a restrained starting point. If petal variation returns, the previous edit was suppressing information that the file could still render.
White balance is also part of the interpretation. Adobe distinguishes white-balance metadata in a raw file from a white balance already applied to JPEG pixel data. Changing a raw rendition can therefore be a different operation from trying to undo a baked-in result. It should aim to represent the illumination and subject appropriately, not merely to force a troublesome graph away from its endpoint.
If the original capture lacks the detail, no slider can reconstruct the exact lost distinctions. Software may infer or synthesize a plausible appearance, but that is different from recovering measurements that were never stored. Raw is useful latitude, not unlimited insurance against exposure errors.
A flower-sized check is more useful than a perfect graph
For a stationary colorful subject, make a restrained comparison rather than chasing a universal histogram shape. Check the RGB channels, capture a slightly lower exposure if circumstances permit, and compare the relevant petals. Keep the comparison on the same display with the same processing so that a screen change does not become another variable.
Lower exposure has costs. It can reduce the signal in dark parts of the scene, so a blanket rule to darken every colorful photograph is poor advice. The point is to protect the detail you value while understanding the tradeoff. A small specular reflection may be expendable; the texture across the main flower may be the reason you took the picture.
When exporting, inspect the final rendition again. A satisfactory raw edit does not guarantee that every later adjustment or output conversion preserves the same component relationships. Equally, an edge spike alone does not tell you whether the lost variation is visible at the size where the photograph will be used.
The best question is narrower than whether the whole photograph is too bright: does the important colored area still contain the distinctions that describe it? Separate channel graphs help you ask that question before a vivid flower becomes a smooth red cutout.