A cross resists generalizations because a cross doesn't run true to type - you can have a Perch/TB cross that comes out looking almost entirely like the Perch parent, entirely like the TB parent, or a combination of the worst traits of both. Or you can have a happy result and get a Perch/TB cross that has the warmblood type conformation that's ideal for dressage, but that's statistically less common. (In the previous thread, a poster get very upset at the criticism of draft crosses as sport horses and post photos of some gorgeous 50% draft crosses that looked and performed like warmbloods. Gorgeous, but not typical.)
So you'd be better off analyzing the confo of the individual horse for suitability for dressage rather than trying to make a generalization about the cross.
Also, you need to consider exactly what you want the horse to do - almost any sound horse can do training and first level, however, if you want to show and be competitive, it's best to have a horse that looks the part and has the movement. If you have aspirations above second level, you must evaluate the horse for its ability to collect.