Yes, you are wrong. The 10/14 ratio applies to the recorded time, not the dead time.
dead_time = 1/(update_rate) - record_length
Which after correcting for your erroneous 23% bonus in the originally presented waveform rates we end up calculating the Rigol has a blind time averaging a respectable 5 times longer than the competing (and more expensive) Agilent model.
My original assertion that the number of divisions decreases the blind time is correct - but the percentages are wrong. As you mentioned, the percentage is the
increase in acquisition time, with a correspondingly smaller (depending on the initial ratio) decrease in blind time.
Rigol's blind time percentage, with a 50,000 wfrm update rate of a 280ns acquisition time (14 divisions * 20ns), is 98.6%
Agilent's blind time percentage, with a 54,000 wfrm update rate of a 200ns acquisition time (10 divisions * 20ns), is 98.9%
Agilent's blind time percentage, if it had a 70,000 wfrm update rate of a 200ns acquisition time, would be 98.6% - exactly what Rigol's is at that timebase.
Of course, as alm pointed out, this does not mean, for example, that the Rigol can be respond to 70k trigger events at that timebase setting - it just means, given an equivalent timebase setting and wfrm rate, that it's less blind than a 10 division scope.
I will re-edit the original post to reflect this data later today.