Well, if you plot a curve through your plain bearing data and another curve through your ball bearing data, it looks to me like the ball bearing curve will be about 20% below the plain bearing curve. In log-log space you'd see fairly straight lines instead of curves.

But you only have three ball bearing data points, so you can't make a firm conclusion. And based on your test setup, a 10-20% difference is probably just above the limit of detectability anyway.

Your point that line diameter and sheave diameter are the dominant factors seems pretty valid. However, I don't think you have good support for saying "there is hardly any difference between plain bearing blocks and ball bearing blocks". The fact that the 47mm ball bearing block outperforms the 58mm plain bearing block indicates that the opposite may be true.