Both of your last two images show the reduction of the alkene after complete reduction of the carbonyl functionality. If that were the case, the independently prepared aldehyde and ester should react the same, yes? So the mechanisms must diverge before complete reduction of the carbonyl functionality to the alkoxide.
I agree that my suggestion maybe or is completely wrong. Now that you have posted some data and even Marshall's (partial) suggestion, it is clear that I don't have a clue.
Eh, aluminum hydride is too hard of a base to do 1,4-reduction. The authors do not propose formation of the saturated alcohol through a 1,4-reduction of the unsaturated ester... but they do invoke the unsaturated ester as part of the mechanism for the formation of the undesired saturated alcohol.
In Table II (ex 2), Marshall suggests that up to 91% of the product comes from 1,4-addition. Isolation of the saturated aldehyde convinces me the enolate is a likely intermediate.
If 1,4-addition is occurring as indicated by Marshall, then saturated alcohol from the aldehyde does not make sense, Table II. Why shouldn't it also result in the saturated aldehyde?
I believe that Marshall was skeptical of these results also. It appeared that post reduction reactions were taking place. The aldehyde was best isolated by inactivating the LAH or limiting LAH available during work up, Table I (ex 5 & 6). That is, looking at the work up conditions, I argue that different product ratios indicate that further reaction takes place between the initial reflux and isolation of the product.
I agree the ester may not have been the primary intermediate. However, a simple 1,4-addition, and work up should release the ester. A 1,4-addition to the unsaturated aldehyde gives the enolate, the enolate goes to aldehyde. If reduction continues during work up, the aldehyde is reduced to the alcohol. While that doesn't completely explain ex 2 & 3 of Table II, but does go in the right direction. If reductant is depleted, then less saturated alcohol and more aldehyde is isolated. When it isn't the opposite. However, since the work up is depleting the hydrides available, a mixed result is found.
I didn't look this up, but will tri-t-butoxyaluminum hydride do a conjugate addition to an aldehyde? It doesn't reduce aldehydes. That was the reason I was hesitant to decisively say what would or wouldn't happen. A big excess of LAH is used and a number of alkoxides are generated to give new reductants. While LAH may not give saturated aldehyde to a great extent, the allyl alcohols can generate alkoxyaluminum hydrides that could be responsible for the small amount of conjugate addition that may have been responsible for the saturated alcohol from the aldehyde. Marshall did not add an equivalent amount of ethanol to the reaction to see what effect the intermediate hydrides may have had on the reduction. However, it seems safe to say that LAH will react faster with an aldehyde than mono-, di- or trialkoxyaluminum hydrides would. Hey, its only a guess.