"Unreported non-event engine failures on a twin" does not categorically change the reported risk profile of a twin vs. single, based on flight hours. You are merely describing an aspect that explains why twins have a different distribution of accident types as compared to a single. There is simply no way to totally understand the risk profile of a twin vs. single without having some understanding of basic data analysis. To explain the differences, you need to look at the breakdown (distribution) of accidents per type, and you'd probably need to control for pilot skill as a lurking explanatory variable.
The Avweb link posted earlier gives us a starting point:
"[in a twin] a variety of other non-weather-related causes are quite significant: botched takeoffs and landings, controlled flight into terrain, improper IFR procedures, fuel exhaustion, and gear-up landings, just to name a few."
"A single is about two-and-a-half times more likely to have an accident due to engine/prop failure than a twin (8% versus 3%)"
"[in a twin] your risk profile changes somewhat... you're less likely to be hurt by an engine failure, and more likely to be victimized by something else."
Seems pretty straightforward: given the current population of pilots and planes, twins are more likely to survive a single engine-out, but twins are also just generally more complicated planes and that complication increases the accident rate. Now if twins also have elevated rates of flight into terrain and IFR mishap, those categories seem to be pilot error that is mostly independent of the equipment, suggesting that twins tend to be flown by pilots that are engaging in flights that require higher proficiency and have a larger risk profile.
So really, we can't get to the complete bottom of this unless we look even deeper at additional data: we need to have training,flight hours, and perhaps some sort of measure of specific pilot expertise for the activity he/she was performing at time of accident. We also need a numerical measure that semi-objectively classifies plane complexity. By regressing on each of these lurking variables, we could then generate a more objective predictor of risk for a given pilot/plane combination. Find the data for me and I'll do it!
Without the data, I think you can only say the following:
singles are probably being flown more conservatively than twins (biases overall safety towards singles)
twins are probably being flown with pilots with greater expertise (bias: twins)
twins have more complicated equipment (bias: singles)
twins are being flown on more difficult missions (bias: singles)
twins have a lesser risk of accident from engine failure (bias: twin)
If you are an expert pilot that flies conservatively in good conditions, you are probably safer in a twin. If you are a ****ty pilot that likes to push it, you are probably safer in a single.
