Before or after taxes is not relevant. If, for that newspaper example, the newspaper's actual P/E (after-taxes) was 30 but Buffett's calculated "intrinsic" P/E was 25, then Buffett would determine that the newspaper was overvalued. If the actual P/E (after taxes) was 20, then he'd determine the paper was undervalued. Or you could use the pre-tax P/E values instead to determine whether the paper was over or under-valued. If the paper's actual pre-tax P/E was above 16, then Buffett would determine the paper was overvalued; if the paper's actual pre-tax P/E was below 16, then Buffett would say undervalued. Whether you use the after-tax or the pre-tax P/E is not relevant. Whichever you use, you want to compare that to a company's actual after-tax or before-tax P/E to determine whether the company is over- or under-valued. That was the point of that example.
Look at Coke. It's P/E is around 50. Its pre-tax P/E will, of course, be lower because EPS will be higher. But whether you use the after-tax P/E or the pre-tax P/E, you want to compare that to Coke's "intrinsic" P/E (what the P/E SHOULD be), as determined by discounting Coke's cash flow.
The point of the newspaper example was not so much to say anything about using after-tax vs. before-tax figures but to show what the change in intrinsic value would be if you change the assumptions used in the perpetual annuity formula. If you say growth is 6% forever, then intrinsic value = $1 mil. / (0.10 - 0.06) = $25 million. $25 mil. would be the appropriate amount to pay if g = 6%. If the company's market capitalization is higher than $25 mil., then it's overvalued. On the other hand, if there is no growth over time but earnings "bob around" around an unchanging value, g = 0% and intrinsic value = $1 mil. / (0.10 - 0.00) = $10 mil. In this case, if the company's market cap. is $15 mil., the company is overvalued.