I might be too late but did you try DyRoBeeS software.
http://www.dyrobes.com/index.htm I am using this software for last 15 years and I had great success in analyzing the branched systems. It is simple and acurate.
I would think that any load is a good point to measure the vibration as long as this load is repeated each time that the vibration is collected. However if this is not possible the "random" collected data trend will ultimately establish a Dead band (high and low overall vibration) levels...