It is worth making some methodological and presentational observations:
Firstly graphs are appealing ways to attract the reader's attention but need care. Yours are log graphs, which flatten the trends to be more linear - that's sort of the point of log linear graphs.
Second reported IRRs are poor measures indeed, DPI or TVPI are almost always needed to make sense of the data. It is worth saying that in even the most comprehensive data sets, over 25% of the data is missing.
Third you don't deal with fund leverage, a massive issue when looking at IRR/DPI/TVPI trends in PE these days. Apple & Pears and all that.
Fourth you only look at LPs returns in funds, not in co-invest or secondary trades. low cost direct investing and portfolio balancing are more available these days, and.;
Finally, you note that the results are net of fees, but do not take the logical step of pointing out that PE is a high fee business that still at least matches a risky small cap portfolio after costs. Therefore the investment strategy is out performing the market, but most of the out performance within a traditional fund is being paid to the GP. That is a pricing problem, not a performance one.

The data has been crawled over many, may times. It always says the same thing: Gross performance is, on average, better than quoted comparable companies (or indeed unquoted comparable companies). Net performance is about the same or a bit better. This just shows that the fund manager captures most of the value they seem to create.