using intraday for the historical high and closing price for the historical low in the price/(hi+lo in last year) formula
is better than intraday for both, closing prices for both, or inverse. Adds quite a bit to CAGR

Is this even reasonably actionable? The sites that publish this information seem to all report intraday highs. You'd have to either find a site that publishes the "wrong" lows or do it yourself by grabbing a year's work of daily close prices for the Nas100 stocks.

At this stage I think we are deep into false accuracy and data mining here. GTR1 gives the range of CAGRs for the 20 cycles of top 5 as 24.8% to 31.3%. That is a 25% wide range.
The additional CAGR using these High and Low figures could well be just a shift in the exact sequence of returns within this range. Probably is, IMHO.

The CAGRs for 3 cycles starting only ONE DAY apart are 26.9%, 28.6%, and 24.8%. So you can get a 2% difference in CAGR just with a one day difference in start date. Not to mention the mismatch between the 20 day holding period in GTR1 vs. the 1 month in a real world portfolio. The backtest doesn't line up with real world.

Little things like that have such a large variance that you can't ascribe any accuracy to the return, it's just in the "see lightning, hear thunder" arena.

I'm not sure how much juice is left in this orange. I don't think we should call the CAGRs to be anything more exact than 2x%