He means 70 degrees of rotation within the daily cycle, so roughly 5 hours of rotation. (He's not talking about the rotational axis orientation being off by 70 degrees.)

5 hours over thousands of years is known to be within the margin of unpredictable irregularities. Nowadays we compensate for these irregularities by altering our timekeeping by way of leap seconds.

I got this idea late, thanks. My mistake. That makes sense.

Those were the numbers I saw when I was looking into this a few weeks back.

If you think about it, a few millennia corresponds to ~1,000,000 days. So to know the rotation angle to within 10 30 degrees or so, you need to know the length of the day to better than one part in 10^7. But each year the length of the day changes by a few tenths of a millisecond, which is about 3 x 10^-8 of the length of a day. So you have to know the change in the length of the day reasonably well not only today, but at all times going back 2000 years.

Those few tenths of a millisecond add up over thousands of years!

Ah, I see where your number comes from. However, those deviations should happen in both senses (adding and subtracting)? But thanks for the computations!

They do but in the long run because of the tidal forces of the moon days become inexorably longer.

Wow, thanks for the details!