It is an interesting balance...and not as cut and dried as one side or the other would like to make out.
Over the years I have lost/accidentally damaged/worn out more game CDs than I care to remember. I'm guessing it is something on the order of seven or eight out of 200+ games (I had something of an addiction when I was younger
)
At the same time I haven't really been in danger of losing the whole lot in one fell swoop as appears to be possible (although clearly not probable) with Steam. Well other than catastrophic house fire, theft, flood or similar rare events...
It is not dissimilar to other cases where a few incidents that cause a large loss get more attention than many many incidents that cause small loss even though the sum of the small incidents far outweighs the sum of the large. It is human nature to focus on the larger incidents as they appear more significant (and blame can be attached to someone...we do love a good blamefest), whereas in reality the many small losses should garner attention.
The best analogy I can come up with (and yes, I realize this is on a completely different level in terms of its real world importance) relates to accidents...many thousands of people die on the roads every year but a handful of plane crashes overshadows them all, to the point that some people refuse to fly, even though the statistics clearly show air travel is far safer mile for mile.
I would love to see the actual statistics (totals) for lost access to purchased games through loss/damage/wearing out of physical media vs. loss of access to games through lost/suspended Steam accounts (normalized for difference in market size of course). My instincts tells me the numbers won't even be close, but maybe your instincts will be different.
Of course those statistics are unlikely to ever be available in a verifiable way so all that remains is guesswork, speculation and the ever popular anecdote.
Does Steam provide more insurance against loss or less?
Frankly I don't see how anyone can tell for sure...but I guarantee opinions will be varied, and most of them will be perfectly valid.
Risk assessment in the absence of reliable statistics is a very personal thing.