I've been struggling with whether to answer this or not. I think I can do so without getting anyone in legal trouble.
Years ago, I noticed this same thing, and I asked a Kenzerco employee essentially this same question at Origins. I will not say what year or if this informant is a current or former employee of Kenzerco. They are a small group and any hint as to their identity would risk exposure. This person hinted to me that to answer my question would be a violation of their non-disclosure agreement.
In any case, as some of you may be aware, Kenzerco do lots of original field research when writing their games. With Strength in Hackmaster 5e this was also the case, or so my informant claimed.
They have a fanatical following on their forums, and one of their forum members was cousins/friends with Jouko Ahola. So they got this forum member to get Jouko to agree to perform a series of strength tests devised by David Kenzer and his cohorts. Ahead of time, the Kenzerco team determined that Jouko would be Strength 20, and then they analyzed the videos of the Finn doing the tests. From this analysis, they got the lifting weight you see.
They did this same strength test to Steve Johansson, a feeble Kenzerco employee. The development team knew Johansson would represent the low end of the scale. They derived all the other values by scaling them between Johansson and Jouko. However, Johansson was so weak he threw the numbers off at the high end, so that accounts for the sharp uptick you see starting at 15. Those numbers they had to fudge!
The weird wobble you see between 2 and 3 is the real life result of Johansson's test.