I thought I'd find the data showing today's valuation is significantly higher than its peak in 1929, but the data I see has today's CAPE ratio at 32.2, and a peak 1929 CAPE ratio of 32.6 occurring in September.

On my figures, using weekly data, the current cap-weight US equities valuation level is 16.2% more expensive than anything from my start of data (1916) through to the 1990s.
The very earliest data get a bit iffy, of course, as there weren't all that many stocks with data available, and I had to splice some different data sources with longer histories. But from around 1926-1930 the market returns data set is better than the one used by most academic researchers.

The difference is probably just that I just use a smoother smoothing method than the E10 in CAPE.
I found four smoothing methods that seem to work adequately, one of which is E10. Then I scaled each of those by whatever factor made it best match current earnings levels best (since each smoothing method has some time lag). Then I take a simple average of those four methods. Why do I do that? E10 has the disadvantage of reporting a sudden upswing in the reported trend earnings on the 10th anniversary of any earnings recession. A "weighted moving average" (WMA) works better, because older data disappear gradually rather than suddenly.

Other smoothing and scaling methods will give different numerical results, but FWIW: My data set shows a current cyclically adjusted earnings yield of 3.251% (analogous to a CAPE of 30.76), versus a 50 year average cyclically adjusted earnings yield of 6.259% (analogous to a CAPE of 15.98). The 25-year valuation (just long enough to catch the tech bubble, credit bubble, and post-pandemic bubbles) gives an average CAEY of 4.067% (analogous to a CAPE of 24.59)--some optimists might consider that the "modern normal", which current levels exceed by only 25%.

Jim