diff --git a/pcsx2/GS/Renderers/HW/GSRendererHW.cpp b/pcsx2/GS/Renderers/HW/GSRendererHW.cpp
index d42230fe34..83c7b5b61b 100644
--- a/pcsx2/GS/Renderers/HW/GSRendererHW.cpp
+++ b/pcsx2/GS/Renderers/HW/GSRendererHW.cpp
@@ -466,18 +466,19 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba, GS
 	if (PRIM->FST)
 	{
 		GL_INS("First vertex is  P: %d => %d    T: %d => %d", v[0].XYZ.X, v[1].XYZ.X, v[0].U, v[1].U);
-
+		const int reversed_pos = (v[0].XYZ.X > v[1].XYZ.X) ? 1 : 0;
+		const int reversed_U = (v[0].U > v[1].U) ? 1 : 0;
 		for (u32 i = 0; i < count; i += 2)
 		{
 			if (write_ba)
-				v[i].XYZ.X -= 128u;
+				v[i + reversed_pos].XYZ.X -= 128u;
 			else
-				v[i + 1].XYZ.X += 128u;
+				v[i + 1 - reversed_pos].XYZ.X += 128u;
 
 			if (read_ba)
-				v[i].U -= 128u;
+				v[i + reversed_U].U -= 128u;
 			else
-				v[i + 1].U += 128u;
+				v[i + 1 - reversed_U].U += 128u;
 
 			if (!half_bottom_vert)
 			{
@@ -503,18 +504,20 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba, GS
 	{
 		const float offset_8pix = 8.0f / tw;
 		GL_INS("First vertex is  P: %d => %d    T: %f => %f (offset %f)", v[0].XYZ.X, v[1].XYZ.X, v[0].ST.S, v[1].ST.S, offset_8pix);
+		const int reversed_pos = (v[0].XYZ.X > v[1].XYZ.X) ? 1 : 0;
+		const int reversed_S = (v[0].ST.S > v[1].ST.S) ? 1 : 0;
 
 		for (u32 i = 0; i < count; i += 2)
 		{
 			if (write_ba)
-				v[i].XYZ.X -= 128u;
+				v[i + reversed_pos].XYZ.X -= 128u;
 			else
-				v[i + 1].XYZ.X += 128u;
+				v[i + 1 - reversed_pos].XYZ.X += 128u;
 
 			if (read_ba)
-				v[i].ST.S -= offset_8pix;
+				v[i + reversed_S].ST.S -= offset_8pix;
 			else
-				v[i + 1].ST.S += offset_8pix;
+				v[i + 1 - reversed_S].ST.S += offset_8pix;
 
 			if (!half_bottom_vert)
 			{
@@ -2627,8 +2630,8 @@ void GSRendererHW::Draw()
 				m_index.tail = 2;
 			}
 		}
-
-		if (rt && (!is_possible_mem_clear || rt->m_TEX0.PSM != FRAME_TEX0.PSM))
+		const bool blending_cd = PRIM->ABE && !m_context->ALPHA.IsOpaque();
+		if (rt && ((!is_possible_mem_clear || blending_cd) || rt->m_TEX0.PSM != FRAME_TEX0.PSM))
 		{
 			if (rt->m_TEX0.TBW != FRAME_TEX0.TBW && !m_cached_ctx.ZBUF.ZMSK && (m_cached_ctx.FRAME.FBMSK & 0xFF000000))
 			{
diff --git a/pcsx2/GS/Renderers/HW/GSTextureCache.cpp b/pcsx2/GS/Renderers/HW/GSTextureCache.cpp
index 18a4e1dcd3..6668defdeb 100644
--- a/pcsx2/GS/Renderers/HW/GSTextureCache.cpp
+++ b/pcsx2/GS/Renderers/HW/GSTextureCache.cpp
@@ -1895,7 +1895,8 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(GIFRegTEX0 TEX0, const GSVe
 						continue;
 
 					dst->m_was_dst_matched = true;
-
+					dst->m_TEX0.TBW = dst_match->m_TEX0.TBW;
+					dst->UpdateValidity(dst->m_valid);
 					if (!CopyRGBFromDepthToColor(dst, dst_match))
 					{
 						// Needed new texture and memory allocation failed.
@@ -1957,8 +1958,12 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(GIFRegTEX0 TEX0, const GSVe
 		{
 			Target* t = *i;
 			// Don't pull in targets without valid lower 24 bits, it makes no sense to convert them.
-			if (bp != t->m_TEX0.TBP0 || !t->m_valid_rgb)
+			// FIXME: Technically the difference in size is fine, but if the target gets reinterpreted, the hw renderer doesn't rearrange the target.
+			// This does cause some extra uploads in some games (like Burnout), but without this, bad data gets displayed in games like Transformers.
+			if (bp != t->m_TEX0.TBP0 || !t->m_valid_rgb || (!is_shuffle && t->m_TEX0.TBW < TEX0.TBW && possible_clear))
+			{
 				continue;
+			}
 
 			const GSLocalMemory::psm_t& t_psm_s = GSLocalMemory::m_psm[t->m_TEX0.PSM];
 			if (t_psm_s.bpp != psm_s.bpp)