Any firm is going to put out polls that don't perform well, and it is impossible to know which those will be until after the election. This is especially the case in the midterms, where turnout is much more unpredictable. My point is that if you look at PPP's full track record across multiple cycles, they stand up to scrutiny fairly well, much better than Gravis Marketing. If you look at their metrics on 538's pollster rankings, they have a very low mean reverted bias (which means they don't tend to miss in the same direction) and a relatively low average error. You chose to cite a handful of the 411 polls in the database, unsurprisingly the handful that were the worst. It paints an inaccurate picture to say the least.
Some Gravis polls are just ridiculous, yeah. But contrary to PPP Gravis doesn’t have a partisan bias while PPP almost always over estimate democrats.