From my experiences over the last year or so for OC dives between 45-85m with up to 90 minutes of deco using GUE ascent standards with also a Buhlmann computer strapped on, the computer normally clears 0-10 minutes before the original plan. (Personally I'm totally happy to accept maybe 10 minutes extra time in the water in return for computer failure being a total non-event.)
(Please do note, GUE ascent standards are NOT the same as UTD RatioDeco (TM), I do not speak for that.)
You know something.... I think I know what's happening here.
When I think back to when I first started technical diving, I dove on tables with a computer (a vytec) that I put in gauge mode to use as a bottom timer. I was cutting tables using Vplanner (I know... but we all did it back then... admit it!) and learning about RD, deep stops and how bad computers were at everything except getting you killed. It was in this period of time that I also picked up the habit of ascending at 3m/min (max) in the shallow zone (after the 50% switch).
The first time I remember ever using a "live" computer during a dive I was diving on a wreck in the North Sea. The dive was to 36m and I was using ean32 and ean50.
I forgot to put the computer in gauge mode before the dive. I noticed it during descent but decided to keep on diving anyway.
The bottom profile was about 36m-50min@ean32 and about 35min of hang time IIRC. I had made quite a few dives like this already so I was pretty sure what to expect. Wreck dives are pretty easy to control in this sense.
Because the computer was turned on it gave me a run time of about ... something like 1:45 min. (it was a while ago but this is probably in the ball park). The run time according to tables was about 90 minutes. I followed my plan and bent the computer but then got curious about what it would have done if I had switched gasses. I had never used the vytec in computer mode during a technical dive up to that point. I bought it for that but then got sucked into the rhetoric online.....
So a few weeks later I was diving on the same wreck. Same protocol, same gasses but this time I put the computer in computer mode and turned on the deco gas. I made a similar profile but this time I changed gasses on the computer and
it did, indeed, clear earlier than my "tables".
Then I got to thinking about why....
At first, I thought what you did..... That the computer and the table dive could be compared on equal terms.
But then I remembered the one most important thing about computers.... they give you longer bottom times because they eliminate the rounding errors built into the tables! We all preach that to our students now. Buy a computer for longer bottom times... nobody makes a bucket profile.....
Even though this was a wreck dive I clearly didn't spend 100% of my time at the maximum depth and the computer knew that. The table was based on an erroneous assumption that this would be the case. The deco times were different because (aside from differences in deco model) the computer knew what I had ACTUALLY done... I, however, did not.
Even calculating average depth on the fly you're going to be off. You have a data point every now and then but the computer samples your depth every few seconds and recalculates the entire dive based upon a very accurate average depth. A human could never come close to that. My conclusion, after thinking it all through, is that, like tables, the average depth calculation has enough "slop" in it to have a significant effect on your deco times.
Maybe not enough to get you bent (on every dive) but enough that trusting in-your-head-average-depth-calculations isn't sounding wise to me when you can strap something to your wrist that does it perfectly.
Just a thought.
R..