That conversion is just ridiculous! There are several ways of looking at it - if a child got a 2 at end EYFS, then they met their end of year expectations, in which case, they could reasonably be expected to get a 2B at the end of Y2. (and nationally this year, 70% of Y2 pupils got 2B, so this would be along the right lines to your 3/4 getting a 2). I would then be looking at how many exceeded at end EYFS, and use this as a guide for a L3 target which is also exceeding the expected level (nationally 24% in 2014).
That said, if, as another poster indicated, the teaching in EYFS has been less than ideal, or if the children came in low and need more than 10 months to catch up with their national peers, then just using end EYFS results might limit your end Y2 targets. What I would do in this scenario is to identify the 25% most able pupils, and target them for L3, then identify the 75% most able to get level 2B+, and 90% to get 2C+. (I stress that this is only if the end EYFS results would suggest a lower attainment than this. If they suggest higher, e.g. if 40% had achieved a 3 at end EYFS, then 40% could be predicted a L3).
What it boils down to is that we use the end EYFS results to predict and target end Y2, but we would never target a cohort of children for results that would be lower than national average. We would of course target for higher than national average if the end EYFS results were suggestive of this, or if historically we tended to achieve above national because of the outstanding progress made in KS1.
But having 3/4 of Y2 children targetted a L3 just doesn't seem right - you would be off the scale! Find out what percentage have got a L3 for the last 3 years, and ask your SLT why the sudden jump?!