Most route-based vibration analysis is performed at the bearing cap locations. For trending, this is the most logical place to detect rotor and bearing faults. Keep in mind that the location that the vibration data is collected may not be the location of "highest" vibration. It also may not indicate the source of the problem.
As an example, many large feed water pumps have oil lubrication lines that can be excited into resonance during operation. When measured at the bearing locations, the vibration may increase to alarm levels. However, at the line itself, the vibration may be orders of magnitude higher. It's very frustrating to be trying to solve a problem when you don't have a good overall sense of the potential problem source.
Alternately, the measurement location(s) specified for trending may be the worst choice for indicating overall response levels. An example of this was a vertical pump system in a nuclear power plant that had a quarterly measurement location specified on the motor end bell that was resonant. The end bell was responding at much higher levels than a more stable housing location just inches away.
In each of these cases, a little investigation beyond the typical trending locations would have indicated the problem source. When in doubt, use your sense of touch and "walk" the machine, its mounts and foundations for changes in vibration level. Then take a few off-route measurements to get a better perspective. It might be time well spent.