Blame the Buckets
The global temperature graph is a combination of temperatures taken on land and at sea. When the researchers looked at the land measurements only they saw no dip in 1945. Records taken at sea, however, do show the dip. This suggests that it may be an artefact of how such measurements were taken during the second world war.
Prior to the war, UK and US fleets had contributed roughly equally to the global temperature record. From 1942 to 1945, UK ships, previously an important gatherer of sea water temperatures, were mobilised on the front and contributed just 5% of measurements. Measurements taken from US ships made up 80%.
The key, then, is how each nation took its measurements. UK ships tended to throw a bucket overboard and lift it on deck to take the water's temperature. US ships by and large would sample water drawn into the engine room before it was used to cool the machinery.
Researchers have known for some time that each method has a bias. Temperatures measured in the buckets tend to be lower than those obtained when a thermometer is placed directly into the ocean because heat escapes from it as it is heaved on deck. The type of bucket can influence the temperature as well: wooden buckets, common in the 19th century, offer better insulation than the canvas buckets used in the 20th century. Engine room measurements, on the hand, tend to be higher than the actual water temperature because these rooms are hot.
So a temperature record dominated by US measurements in the early 1940s would show the sea surface to be warmer than it actually was at the time.
Moreover, late in 1945, the UK resumed its measurements and for a period was responsible for half the global record while the US share dropped to 30%. This period is biased towards cooler, bucket-based temperatures, and corresponds to the sudden 1945 dip.