I printed the E2 version of this and tried to test feed some filament through it before hooking everything up just as a function check. I didn't work out so great. The sample filament I cut off to test with was from the inner part of a roll so it has more curl than filament from the middle or outer part of a roll. The problem is the open feed path under the ball is long enough for the curl to reform and change the filament direction enough to cause the feed to fail in most orientations. The only orientation that works is towards the magnet. If it the curl faces towards the window or towards the ball, it will fail to load. Note filament from the middle or outer part of a spool works OK. All samples were cleanly cut with flush cutters.
The root cause of the problem appears to be with the Prusa design. It appears this 7mm ball bearing size was chosen because the FINDA probe is 7.88mm wide at the threads and they just continued the hole down, resulting in a long free section of the feed path. I'm wondering if it wouldn't be better to narrow the bottom of the opening down for a 4mm bearing, which are readily available. This would reduce the "free" section of the feed path significantly and therefore the chance the filament would get stuck on feeds if it isn't straight enough.